git/builtin/name-rev.c

687 lines
17 KiB
C
Raw Normal View History

#include "builtin.h"
#include "cache.h"
#include "repository.h"
#include "config.h"
#include "commit.h"
#include "tag.h"
#include "refs.h"
#include "parse-options.h"
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
#include "prio-queue.h"
#include "hash-lookup.h"
#include "commit-slab.h"
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
#include "commit-graph.h"
/*
* One day. See the 'name a rev shortly after epoch' test in t6120 when
* changing this value
*/
#define CUTOFF_DATE_SLOP 86400
struct rev_name {
const char *tip_name;
timestamp_t taggerdate;
int generation;
int distance;
name-rev: favor describing with tags and use committer date to tiebreak "git name-rev" assigned a phony "far in the future" date to tips of refs that are not pointing at tag objects, and favored names based on a ref with the oldest date. This made it almost impossible for an unannotated tags and branches to be counted as a viable base, which was especially problematic when the command is run with the "--tags" option. If an unannotated tag that points at an ancient commit and an annotated tag that points at a much newer commit reaches the commit that is being named, the old unannotated tag was ignored. Update the "taggerdate" field of the rev-name structure, which is initialized from the tip of ref, to have the committer date if the object at the tip of ref is a commit, not a tag, so that we can optionally take it into account when doing "is this name better?" comparison logic. When "name-rev" is run without the "--tags" option, the general expectation is still to name the commit based on a tag if possible, but use non-tag refs as fallback, and tiebreak among these non-tag refs by favoring names with shorter hops from the tip. The use of a phony "far in the future" date in the original code was an effective way to ensure this expectation is held: a non-tag tip gets the same "far in the future" timestamp, giving precedence to tags, and among non-tag tips, names with shorter hops are preferred over longer hops, without taking the "taggerdate" into account. As we are taking over the "taggerdate" field to store the committer date for tips with commits: (1) keep the original logic when comparing names based on two refs both of which are from refs/tags/; (2) favoring a name based on a ref in refs/tags/ hierarchy over a ref outside the hierarchy; (3) between two names based on a ref both outside refs/tags/, give precedence to a name with shorter hops and use "taggerdate" only to tie-break. A change to t4202 is a natural consequence. The test creates a commit on a branch "side" and points at it with an unannotated tag "refs/tags/side-2". The original code couldn't decide which one to favor at all, and gave a name based on a branch (simply because refs/heads/side sorts earlier than refs/tags/side-2). Because the updated logic is taught to favor refs in refs/tags/ hierarchy, the the test is updated to expect to see tags/side-2 instead. [mjg: open-coded the comparisons in is_better_name(), dropping a helper macro used in the original] Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@grubix.eu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-03-29 14:39:16 +00:00
int from_tag;
};
define_commit_slab(commit_rev_name, struct rev_name);
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
static timestamp_t generation_cutoff = GENERATION_NUMBER_INFINITY;
static timestamp_t cutoff = TIME_MAX;
static struct commit_rev_name rev_names;
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
/* Disable the cutoff checks entirely */
static void disable_cutoff(void)
{
generation_cutoff = 0;
cutoff = 0;
}
/* Cutoff searching any commits older than this one */
static void set_commit_cutoff(struct commit *commit)
{
if (cutoff > commit->date)
cutoff = commit->date;
if (generation_cutoff) {
timestamp_t generation = commit_graph_generation(commit);
if (generation_cutoff > generation)
generation_cutoff = generation;
}
}
/* adjust the commit date cutoff with a slop to allow for slightly incorrect
* commit timestamps in case of clock skew.
*/
static void adjust_cutoff_timestamp_for_slop(void)
{
if (cutoff) {
/* check for undeflow */
if (cutoff > TIME_MIN + CUTOFF_DATE_SLOP)
cutoff = cutoff - CUTOFF_DATE_SLOP;
else
cutoff = TIME_MIN;
}
}
/* Check if a commit is before the cutoff. Prioritize generation numbers
* first, but use the commit timestamp if we lack generation data.
*/
static int commit_is_before_cutoff(struct commit *commit)
{
if (generation_cutoff < GENERATION_NUMBER_INFINITY)
return generation_cutoff &&
commit_graph_generation(commit) < generation_cutoff;
return commit->date < cutoff;
}
/* How many generations are maximally preferred over _one_ merge traversal? */
#define MERGE_TRAVERSAL_WEIGHT 65535
static int is_valid_rev_name(const struct rev_name *name)
{
return name && name->tip_name;
}
static struct rev_name *get_commit_rev_name(const struct commit *commit)
{
struct rev_name *name = commit_rev_name_peek(&rev_names, commit);
return is_valid_rev_name(name) ? name : NULL;
}
name-rev: prefer shorter names over following merges name-rev has a MERGE_TRAVERSAL_WEIGHT to say that traversing a second or later parent of a merge should be 65535 times more expensive than a first-parent traversal, as per ac076c29ae8d (name-rev: Fix non-shortest description, 2007-08-27). The point of this weight is to prefer names like v2.32.0~1471^2 over names like v2.32.0~43^2~15^2~11^2~20^2~31^2 which are two equally valid names in git.git for the same commit. Note that the first follows 1472 parent traversals compared to a mere 125 for the second. Weighting all traversals equally would clearly prefer the second name since it has fewer parent traversals, but humans aren't going to be traversing commits and they tend to have an easier time digesting names with fewer segments. The fact that the former only has two segments (~1471, ^2) makes it much simpler than the latter which has six segments (~43, ^2, ~15, etc.). Since name-rev is meant to "find symbolic names suitable for human digestion", we prefer fewer segments. However, the particular rule implemented in name-rev would actually prefer v2.33.0-rc0~11^2~1 over v2.33.0-rc0~20^2 because both have precisely one second parent traversal, and it gives the tie breaker to shortest number of total parent traversals. Fewer segments is more important for human consumption than number of hops, so we'd rather see the latter which has one fewer segment. Include the generation in is_better_name() and use a new effective_distance() calculation so that we prefer fewer segments in the printed name over fewer total parent traversals performed to get the answer. == Side-note on tie-breakers == When there are the same number of segments for two different names, we actually use the name of an ancestor commit as a tie-breaker as well. For example, for the commit cbdca289fb in the git.git repository, we prefer the name v2.33.0-rc0~112^2~1 over v2.33.0-rc0~57^2~5. This is because: * cbdca289fb is the parent of 25e65b6dd5, which implies the name for cbdca289fb should be the first parent of the preferred name for 25e65b6dd5 * 25e65b6dd5 could be named either v2.33.0-rc0~112^2 or v2.33.0-rc0~57^2~4, but the former is preferred over the latter due to fewer segments * combine the two previous facts, and the name we get for cbdca289fb is "v2.33.0-rc0~112^2~1" rather than "v2.33.0-rc0~57^2~5". Technically, we get this for free out of the implementation since we only keep track of one name for each commit as we walk history (and re-add parents to the queue if we find a better name for those parents), but the first bullet point above ensures users get results that feel more consistent. == Alternative Ideas and Meanings Discussed == One suggestion that came up during review was that shortest string-length might be easiest for users to consume. However, such a scheme would be rather computationally expensive (we'd have to track all names for each commit as we traversed the graph) and would additionally come with the possibly perplexing result that on a linear segment of history we could rapidly swap back and forth on names: MYTAG~3^2 would be preferred over MYTAG~9998 MYTAG~3^2~1 would NOT be preferred over MYTAG~9999 MYTAG~3^2~2 might be preferred over MYTAG~10000 Another item that came up was possible auxiliary semantic meanings for name-rev results either before or after this patch. The basic answer was that the previous implementation had no known useful auxiliary semantics, but that for many repositories (most in my experience), the new scheme does. In particular, the new name-rev output can often be used to answer the question, "How or when did this commit get merged?" Since that usefulness depends on how merges happen within the repository and thus isn't universally applicable, details are omitted here but you can see them at [1]. [1] https://lore.kernel.org/git/CABPp-BEeUM+3NLKDVdak90_UUeNghYCx=Dgir6=8ixvYmvyq3Q@mail.gmail.com/ Finally, it was noted that the algorithm could be improved by just explicitly tracking the number of segments and using both it and distance in the comparison, instead of giving a magic number that tries to blend the two (and which therefore might give suboptimal results in repositories with really huge numbers of commits that periodically merge older code). However, "[this patch] seems to give us a much better results than the current code, so let's take it and leave further futzing outside the scope." Signed-off-by: Elijah Newren <newren@gmail.com> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-12-04 05:35:52 +00:00
static int effective_distance(int distance, int generation)
{
return distance + (generation > 0 ? MERGE_TRAVERSAL_WEIGHT : 0);
}
static int is_better_name(struct rev_name *name,
timestamp_t taggerdate,
name-rev: prefer shorter names over following merges name-rev has a MERGE_TRAVERSAL_WEIGHT to say that traversing a second or later parent of a merge should be 65535 times more expensive than a first-parent traversal, as per ac076c29ae8d (name-rev: Fix non-shortest description, 2007-08-27). The point of this weight is to prefer names like v2.32.0~1471^2 over names like v2.32.0~43^2~15^2~11^2~20^2~31^2 which are two equally valid names in git.git for the same commit. Note that the first follows 1472 parent traversals compared to a mere 125 for the second. Weighting all traversals equally would clearly prefer the second name since it has fewer parent traversals, but humans aren't going to be traversing commits and they tend to have an easier time digesting names with fewer segments. The fact that the former only has two segments (~1471, ^2) makes it much simpler than the latter which has six segments (~43, ^2, ~15, etc.). Since name-rev is meant to "find symbolic names suitable for human digestion", we prefer fewer segments. However, the particular rule implemented in name-rev would actually prefer v2.33.0-rc0~11^2~1 over v2.33.0-rc0~20^2 because both have precisely one second parent traversal, and it gives the tie breaker to shortest number of total parent traversals. Fewer segments is more important for human consumption than number of hops, so we'd rather see the latter which has one fewer segment. Include the generation in is_better_name() and use a new effective_distance() calculation so that we prefer fewer segments in the printed name over fewer total parent traversals performed to get the answer. == Side-note on tie-breakers == When there are the same number of segments for two different names, we actually use the name of an ancestor commit as a tie-breaker as well. For example, for the commit cbdca289fb in the git.git repository, we prefer the name v2.33.0-rc0~112^2~1 over v2.33.0-rc0~57^2~5. This is because: * cbdca289fb is the parent of 25e65b6dd5, which implies the name for cbdca289fb should be the first parent of the preferred name for 25e65b6dd5 * 25e65b6dd5 could be named either v2.33.0-rc0~112^2 or v2.33.0-rc0~57^2~4, but the former is preferred over the latter due to fewer segments * combine the two previous facts, and the name we get for cbdca289fb is "v2.33.0-rc0~112^2~1" rather than "v2.33.0-rc0~57^2~5". Technically, we get this for free out of the implementation since we only keep track of one name for each commit as we walk history (and re-add parents to the queue if we find a better name for those parents), but the first bullet point above ensures users get results that feel more consistent. == Alternative Ideas and Meanings Discussed == One suggestion that came up during review was that shortest string-length might be easiest for users to consume. However, such a scheme would be rather computationally expensive (we'd have to track all names for each commit as we traversed the graph) and would additionally come with the possibly perplexing result that on a linear segment of history we could rapidly swap back and forth on names: MYTAG~3^2 would be preferred over MYTAG~9998 MYTAG~3^2~1 would NOT be preferred over MYTAG~9999 MYTAG~3^2~2 might be preferred over MYTAG~10000 Another item that came up was possible auxiliary semantic meanings for name-rev results either before or after this patch. The basic answer was that the previous implementation had no known useful auxiliary semantics, but that for many repositories (most in my experience), the new scheme does. In particular, the new name-rev output can often be used to answer the question, "How or when did this commit get merged?" Since that usefulness depends on how merges happen within the repository and thus isn't universally applicable, details are omitted here but you can see them at [1]. [1] https://lore.kernel.org/git/CABPp-BEeUM+3NLKDVdak90_UUeNghYCx=Dgir6=8ixvYmvyq3Q@mail.gmail.com/ Finally, it was noted that the algorithm could be improved by just explicitly tracking the number of segments and using both it and distance in the comparison, instead of giving a magic number that tries to blend the two (and which therefore might give suboptimal results in repositories with really huge numbers of commits that periodically merge older code). However, "[this patch] seems to give us a much better results than the current code, so let's take it and leave further futzing outside the scope." Signed-off-by: Elijah Newren <newren@gmail.com> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-12-04 05:35:52 +00:00
int generation,
name-rev: favor describing with tags and use committer date to tiebreak "git name-rev" assigned a phony "far in the future" date to tips of refs that are not pointing at tag objects, and favored names based on a ref with the oldest date. This made it almost impossible for an unannotated tags and branches to be counted as a viable base, which was especially problematic when the command is run with the "--tags" option. If an unannotated tag that points at an ancient commit and an annotated tag that points at a much newer commit reaches the commit that is being named, the old unannotated tag was ignored. Update the "taggerdate" field of the rev-name structure, which is initialized from the tip of ref, to have the committer date if the object at the tip of ref is a commit, not a tag, so that we can optionally take it into account when doing "is this name better?" comparison logic. When "name-rev" is run without the "--tags" option, the general expectation is still to name the commit based on a tag if possible, but use non-tag refs as fallback, and tiebreak among these non-tag refs by favoring names with shorter hops from the tip. The use of a phony "far in the future" date in the original code was an effective way to ensure this expectation is held: a non-tag tip gets the same "far in the future" timestamp, giving precedence to tags, and among non-tag tips, names with shorter hops are preferred over longer hops, without taking the "taggerdate" into account. As we are taking over the "taggerdate" field to store the committer date for tips with commits: (1) keep the original logic when comparing names based on two refs both of which are from refs/tags/; (2) favoring a name based on a ref in refs/tags/ hierarchy over a ref outside the hierarchy; (3) between two names based on a ref both outside refs/tags/, give precedence to a name with shorter hops and use "taggerdate" only to tie-break. A change to t4202 is a natural consequence. The test creates a commit on a branch "side" and points at it with an unannotated tag "refs/tags/side-2". The original code couldn't decide which one to favor at all, and gave a name based on a branch (simply because refs/heads/side sorts earlier than refs/tags/side-2). Because the updated logic is taught to favor refs in refs/tags/ hierarchy, the the test is updated to expect to see tags/side-2 instead. [mjg: open-coded the comparisons in is_better_name(), dropping a helper macro used in the original] Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@grubix.eu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-03-29 14:39:16 +00:00
int distance,
int from_tag)
{
name-rev: prefer shorter names over following merges name-rev has a MERGE_TRAVERSAL_WEIGHT to say that traversing a second or later parent of a merge should be 65535 times more expensive than a first-parent traversal, as per ac076c29ae8d (name-rev: Fix non-shortest description, 2007-08-27). The point of this weight is to prefer names like v2.32.0~1471^2 over names like v2.32.0~43^2~15^2~11^2~20^2~31^2 which are two equally valid names in git.git for the same commit. Note that the first follows 1472 parent traversals compared to a mere 125 for the second. Weighting all traversals equally would clearly prefer the second name since it has fewer parent traversals, but humans aren't going to be traversing commits and they tend to have an easier time digesting names with fewer segments. The fact that the former only has two segments (~1471, ^2) makes it much simpler than the latter which has six segments (~43, ^2, ~15, etc.). Since name-rev is meant to "find symbolic names suitable for human digestion", we prefer fewer segments. However, the particular rule implemented in name-rev would actually prefer v2.33.0-rc0~11^2~1 over v2.33.0-rc0~20^2 because both have precisely one second parent traversal, and it gives the tie breaker to shortest number of total parent traversals. Fewer segments is more important for human consumption than number of hops, so we'd rather see the latter which has one fewer segment. Include the generation in is_better_name() and use a new effective_distance() calculation so that we prefer fewer segments in the printed name over fewer total parent traversals performed to get the answer. == Side-note on tie-breakers == When there are the same number of segments for two different names, we actually use the name of an ancestor commit as a tie-breaker as well. For example, for the commit cbdca289fb in the git.git repository, we prefer the name v2.33.0-rc0~112^2~1 over v2.33.0-rc0~57^2~5. This is because: * cbdca289fb is the parent of 25e65b6dd5, which implies the name for cbdca289fb should be the first parent of the preferred name for 25e65b6dd5 * 25e65b6dd5 could be named either v2.33.0-rc0~112^2 or v2.33.0-rc0~57^2~4, but the former is preferred over the latter due to fewer segments * combine the two previous facts, and the name we get for cbdca289fb is "v2.33.0-rc0~112^2~1" rather than "v2.33.0-rc0~57^2~5". Technically, we get this for free out of the implementation since we only keep track of one name for each commit as we walk history (and re-add parents to the queue if we find a better name for those parents), but the first bullet point above ensures users get results that feel more consistent. == Alternative Ideas and Meanings Discussed == One suggestion that came up during review was that shortest string-length might be easiest for users to consume. However, such a scheme would be rather computationally expensive (we'd have to track all names for each commit as we traversed the graph) and would additionally come with the possibly perplexing result that on a linear segment of history we could rapidly swap back and forth on names: MYTAG~3^2 would be preferred over MYTAG~9998 MYTAG~3^2~1 would NOT be preferred over MYTAG~9999 MYTAG~3^2~2 might be preferred over MYTAG~10000 Another item that came up was possible auxiliary semantic meanings for name-rev results either before or after this patch. The basic answer was that the previous implementation had no known useful auxiliary semantics, but that for many repositories (most in my experience), the new scheme does. In particular, the new name-rev output can often be used to answer the question, "How or when did this commit get merged?" Since that usefulness depends on how merges happen within the repository and thus isn't universally applicable, details are omitted here but you can see them at [1]. [1] https://lore.kernel.org/git/CABPp-BEeUM+3NLKDVdak90_UUeNghYCx=Dgir6=8ixvYmvyq3Q@mail.gmail.com/ Finally, it was noted that the algorithm could be improved by just explicitly tracking the number of segments and using both it and distance in the comparison, instead of giving a magic number that tries to blend the two (and which therefore might give suboptimal results in repositories with really huge numbers of commits that periodically merge older code). However, "[this patch] seems to give us a much better results than the current code, so let's take it and leave further futzing outside the scope." Signed-off-by: Elijah Newren <newren@gmail.com> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-12-04 05:35:52 +00:00
int name_distance = effective_distance(name->distance, name->generation);
int new_distance = effective_distance(distance, generation);
name-rev: favor describing with tags and use committer date to tiebreak "git name-rev" assigned a phony "far in the future" date to tips of refs that are not pointing at tag objects, and favored names based on a ref with the oldest date. This made it almost impossible for an unannotated tags and branches to be counted as a viable base, which was especially problematic when the command is run with the "--tags" option. If an unannotated tag that points at an ancient commit and an annotated tag that points at a much newer commit reaches the commit that is being named, the old unannotated tag was ignored. Update the "taggerdate" field of the rev-name structure, which is initialized from the tip of ref, to have the committer date if the object at the tip of ref is a commit, not a tag, so that we can optionally take it into account when doing "is this name better?" comparison logic. When "name-rev" is run without the "--tags" option, the general expectation is still to name the commit based on a tag if possible, but use non-tag refs as fallback, and tiebreak among these non-tag refs by favoring names with shorter hops from the tip. The use of a phony "far in the future" date in the original code was an effective way to ensure this expectation is held: a non-tag tip gets the same "far in the future" timestamp, giving precedence to tags, and among non-tag tips, names with shorter hops are preferred over longer hops, without taking the "taggerdate" into account. As we are taking over the "taggerdate" field to store the committer date for tips with commits: (1) keep the original logic when comparing names based on two refs both of which are from refs/tags/; (2) favoring a name based on a ref in refs/tags/ hierarchy over a ref outside the hierarchy; (3) between two names based on a ref both outside refs/tags/, give precedence to a name with shorter hops and use "taggerdate" only to tie-break. A change to t4202 is a natural consequence. The test creates a commit on a branch "side" and points at it with an unannotated tag "refs/tags/side-2". The original code couldn't decide which one to favor at all, and gave a name based on a branch (simply because refs/heads/side sorts earlier than refs/tags/side-2). Because the updated logic is taught to favor refs in refs/tags/ hierarchy, the the test is updated to expect to see tags/side-2 instead. [mjg: open-coded the comparisons in is_better_name(), dropping a helper macro used in the original] Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@grubix.eu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-03-29 14:39:16 +00:00
/*
* When comparing names based on tags, prefer names
* based on the older tag, even if it is farther away.
*/
if (from_tag && name->from_tag)
return (name->taggerdate > taggerdate ||
(name->taggerdate == taggerdate &&
name-rev: prefer shorter names over following merges name-rev has a MERGE_TRAVERSAL_WEIGHT to say that traversing a second or later parent of a merge should be 65535 times more expensive than a first-parent traversal, as per ac076c29ae8d (name-rev: Fix non-shortest description, 2007-08-27). The point of this weight is to prefer names like v2.32.0~1471^2 over names like v2.32.0~43^2~15^2~11^2~20^2~31^2 which are two equally valid names in git.git for the same commit. Note that the first follows 1472 parent traversals compared to a mere 125 for the second. Weighting all traversals equally would clearly prefer the second name since it has fewer parent traversals, but humans aren't going to be traversing commits and they tend to have an easier time digesting names with fewer segments. The fact that the former only has two segments (~1471, ^2) makes it much simpler than the latter which has six segments (~43, ^2, ~15, etc.). Since name-rev is meant to "find symbolic names suitable for human digestion", we prefer fewer segments. However, the particular rule implemented in name-rev would actually prefer v2.33.0-rc0~11^2~1 over v2.33.0-rc0~20^2 because both have precisely one second parent traversal, and it gives the tie breaker to shortest number of total parent traversals. Fewer segments is more important for human consumption than number of hops, so we'd rather see the latter which has one fewer segment. Include the generation in is_better_name() and use a new effective_distance() calculation so that we prefer fewer segments in the printed name over fewer total parent traversals performed to get the answer. == Side-note on tie-breakers == When there are the same number of segments for two different names, we actually use the name of an ancestor commit as a tie-breaker as well. For example, for the commit cbdca289fb in the git.git repository, we prefer the name v2.33.0-rc0~112^2~1 over v2.33.0-rc0~57^2~5. This is because: * cbdca289fb is the parent of 25e65b6dd5, which implies the name for cbdca289fb should be the first parent of the preferred name for 25e65b6dd5 * 25e65b6dd5 could be named either v2.33.0-rc0~112^2 or v2.33.0-rc0~57^2~4, but the former is preferred over the latter due to fewer segments * combine the two previous facts, and the name we get for cbdca289fb is "v2.33.0-rc0~112^2~1" rather than "v2.33.0-rc0~57^2~5". Technically, we get this for free out of the implementation since we only keep track of one name for each commit as we walk history (and re-add parents to the queue if we find a better name for those parents), but the first bullet point above ensures users get results that feel more consistent. == Alternative Ideas and Meanings Discussed == One suggestion that came up during review was that shortest string-length might be easiest for users to consume. However, such a scheme would be rather computationally expensive (we'd have to track all names for each commit as we traversed the graph) and would additionally come with the possibly perplexing result that on a linear segment of history we could rapidly swap back and forth on names: MYTAG~3^2 would be preferred over MYTAG~9998 MYTAG~3^2~1 would NOT be preferred over MYTAG~9999 MYTAG~3^2~2 might be preferred over MYTAG~10000 Another item that came up was possible auxiliary semantic meanings for name-rev results either before or after this patch. The basic answer was that the previous implementation had no known useful auxiliary semantics, but that for many repositories (most in my experience), the new scheme does. In particular, the new name-rev output can often be used to answer the question, "How or when did this commit get merged?" Since that usefulness depends on how merges happen within the repository and thus isn't universally applicable, details are omitted here but you can see them at [1]. [1] https://lore.kernel.org/git/CABPp-BEeUM+3NLKDVdak90_UUeNghYCx=Dgir6=8ixvYmvyq3Q@mail.gmail.com/ Finally, it was noted that the algorithm could be improved by just explicitly tracking the number of segments and using both it and distance in the comparison, instead of giving a magic number that tries to blend the two (and which therefore might give suboptimal results in repositories with really huge numbers of commits that periodically merge older code). However, "[this patch] seems to give us a much better results than the current code, so let's take it and leave further futzing outside the scope." Signed-off-by: Elijah Newren <newren@gmail.com> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-12-04 05:35:52 +00:00
name_distance > new_distance));
name-rev: favor describing with tags and use committer date to tiebreak "git name-rev" assigned a phony "far in the future" date to tips of refs that are not pointing at tag objects, and favored names based on a ref with the oldest date. This made it almost impossible for an unannotated tags and branches to be counted as a viable base, which was especially problematic when the command is run with the "--tags" option. If an unannotated tag that points at an ancient commit and an annotated tag that points at a much newer commit reaches the commit that is being named, the old unannotated tag was ignored. Update the "taggerdate" field of the rev-name structure, which is initialized from the tip of ref, to have the committer date if the object at the tip of ref is a commit, not a tag, so that we can optionally take it into account when doing "is this name better?" comparison logic. When "name-rev" is run without the "--tags" option, the general expectation is still to name the commit based on a tag if possible, but use non-tag refs as fallback, and tiebreak among these non-tag refs by favoring names with shorter hops from the tip. The use of a phony "far in the future" date in the original code was an effective way to ensure this expectation is held: a non-tag tip gets the same "far in the future" timestamp, giving precedence to tags, and among non-tag tips, names with shorter hops are preferred over longer hops, without taking the "taggerdate" into account. As we are taking over the "taggerdate" field to store the committer date for tips with commits: (1) keep the original logic when comparing names based on two refs both of which are from refs/tags/; (2) favoring a name based on a ref in refs/tags/ hierarchy over a ref outside the hierarchy; (3) between two names based on a ref both outside refs/tags/, give precedence to a name with shorter hops and use "taggerdate" only to tie-break. A change to t4202 is a natural consequence. The test creates a commit on a branch "side" and points at it with an unannotated tag "refs/tags/side-2". The original code couldn't decide which one to favor at all, and gave a name based on a branch (simply because refs/heads/side sorts earlier than refs/tags/side-2). Because the updated logic is taught to favor refs in refs/tags/ hierarchy, the the test is updated to expect to see tags/side-2 instead. [mjg: open-coded the comparisons in is_better_name(), dropping a helper macro used in the original] Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@grubix.eu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-03-29 14:39:16 +00:00
/*
* We know that at least one of them is a non-tag at this point.
* favor a tag over a non-tag.
*/
if (name->from_tag != from_tag)
return from_tag;
/*
* We are now looking at two non-tags. Tiebreak to favor
* shorter hops.
*/
name-rev: prefer shorter names over following merges name-rev has a MERGE_TRAVERSAL_WEIGHT to say that traversing a second or later parent of a merge should be 65535 times more expensive than a first-parent traversal, as per ac076c29ae8d (name-rev: Fix non-shortest description, 2007-08-27). The point of this weight is to prefer names like v2.32.0~1471^2 over names like v2.32.0~43^2~15^2~11^2~20^2~31^2 which are two equally valid names in git.git for the same commit. Note that the first follows 1472 parent traversals compared to a mere 125 for the second. Weighting all traversals equally would clearly prefer the second name since it has fewer parent traversals, but humans aren't going to be traversing commits and they tend to have an easier time digesting names with fewer segments. The fact that the former only has two segments (~1471, ^2) makes it much simpler than the latter which has six segments (~43, ^2, ~15, etc.). Since name-rev is meant to "find symbolic names suitable for human digestion", we prefer fewer segments. However, the particular rule implemented in name-rev would actually prefer v2.33.0-rc0~11^2~1 over v2.33.0-rc0~20^2 because both have precisely one second parent traversal, and it gives the tie breaker to shortest number of total parent traversals. Fewer segments is more important for human consumption than number of hops, so we'd rather see the latter which has one fewer segment. Include the generation in is_better_name() and use a new effective_distance() calculation so that we prefer fewer segments in the printed name over fewer total parent traversals performed to get the answer. == Side-note on tie-breakers == When there are the same number of segments for two different names, we actually use the name of an ancestor commit as a tie-breaker as well. For example, for the commit cbdca289fb in the git.git repository, we prefer the name v2.33.0-rc0~112^2~1 over v2.33.0-rc0~57^2~5. This is because: * cbdca289fb is the parent of 25e65b6dd5, which implies the name for cbdca289fb should be the first parent of the preferred name for 25e65b6dd5 * 25e65b6dd5 could be named either v2.33.0-rc0~112^2 or v2.33.0-rc0~57^2~4, but the former is preferred over the latter due to fewer segments * combine the two previous facts, and the name we get for cbdca289fb is "v2.33.0-rc0~112^2~1" rather than "v2.33.0-rc0~57^2~5". Technically, we get this for free out of the implementation since we only keep track of one name for each commit as we walk history (and re-add parents to the queue if we find a better name for those parents), but the first bullet point above ensures users get results that feel more consistent. == Alternative Ideas and Meanings Discussed == One suggestion that came up during review was that shortest string-length might be easiest for users to consume. However, such a scheme would be rather computationally expensive (we'd have to track all names for each commit as we traversed the graph) and would additionally come with the possibly perplexing result that on a linear segment of history we could rapidly swap back and forth on names: MYTAG~3^2 would be preferred over MYTAG~9998 MYTAG~3^2~1 would NOT be preferred over MYTAG~9999 MYTAG~3^2~2 might be preferred over MYTAG~10000 Another item that came up was possible auxiliary semantic meanings for name-rev results either before or after this patch. The basic answer was that the previous implementation had no known useful auxiliary semantics, but that for many repositories (most in my experience), the new scheme does. In particular, the new name-rev output can often be used to answer the question, "How or when did this commit get merged?" Since that usefulness depends on how merges happen within the repository and thus isn't universally applicable, details are omitted here but you can see them at [1]. [1] https://lore.kernel.org/git/CABPp-BEeUM+3NLKDVdak90_UUeNghYCx=Dgir6=8ixvYmvyq3Q@mail.gmail.com/ Finally, it was noted that the algorithm could be improved by just explicitly tracking the number of segments and using both it and distance in the comparison, instead of giving a magic number that tries to blend the two (and which therefore might give suboptimal results in repositories with really huge numbers of commits that periodically merge older code). However, "[this patch] seems to give us a much better results than the current code, so let's take it and leave further futzing outside the scope." Signed-off-by: Elijah Newren <newren@gmail.com> Acked-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Acked-by: Johannes Schindelin <Johannes.Schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-12-04 05:35:52 +00:00
if (name_distance != new_distance)
return name_distance > new_distance;
name-rev: favor describing with tags and use committer date to tiebreak "git name-rev" assigned a phony "far in the future" date to tips of refs that are not pointing at tag objects, and favored names based on a ref with the oldest date. This made it almost impossible for an unannotated tags and branches to be counted as a viable base, which was especially problematic when the command is run with the "--tags" option. If an unannotated tag that points at an ancient commit and an annotated tag that points at a much newer commit reaches the commit that is being named, the old unannotated tag was ignored. Update the "taggerdate" field of the rev-name structure, which is initialized from the tip of ref, to have the committer date if the object at the tip of ref is a commit, not a tag, so that we can optionally take it into account when doing "is this name better?" comparison logic. When "name-rev" is run without the "--tags" option, the general expectation is still to name the commit based on a tag if possible, but use non-tag refs as fallback, and tiebreak among these non-tag refs by favoring names with shorter hops from the tip. The use of a phony "far in the future" date in the original code was an effective way to ensure this expectation is held: a non-tag tip gets the same "far in the future" timestamp, giving precedence to tags, and among non-tag tips, names with shorter hops are preferred over longer hops, without taking the "taggerdate" into account. As we are taking over the "taggerdate" field to store the committer date for tips with commits: (1) keep the original logic when comparing names based on two refs both of which are from refs/tags/; (2) favoring a name based on a ref in refs/tags/ hierarchy over a ref outside the hierarchy; (3) between two names based on a ref both outside refs/tags/, give precedence to a name with shorter hops and use "taggerdate" only to tie-break. A change to t4202 is a natural consequence. The test creates a commit on a branch "side" and points at it with an unannotated tag "refs/tags/side-2". The original code couldn't decide which one to favor at all, and gave a name based on a branch (simply because refs/heads/side sorts earlier than refs/tags/side-2). Because the updated logic is taught to favor refs in refs/tags/ hierarchy, the the test is updated to expect to see tags/side-2 instead. [mjg: open-coded the comparisons in is_better_name(), dropping a helper macro used in the original] Signed-off-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@grubix.eu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-03-29 14:39:16 +00:00
/* ... or tiebreak to favor older date */
if (name->taggerdate != taggerdate)
return name->taggerdate > taggerdate;
/* keep the current one if we cannot decide */
return 0;
}
static struct rev_name *create_or_update_name(struct commit *commit,
timestamp_t taggerdate,
int generation, int distance,
int from_tag)
{
struct rev_name *name = commit_rev_name_at(&rev_names, commit);
if (is_valid_rev_name(name) &&
!is_better_name(name, taggerdate, generation, distance, from_tag))
return NULL;
name->taggerdate = taggerdate;
name->generation = generation;
name->distance = distance;
name->from_tag = from_tag;
return name;
}
static char *get_parent_name(const struct rev_name *name, int parent_number)
{
struct strbuf sb = STRBUF_INIT;
size_t len;
strip_suffix(name->tip_name, "^0", &len);
if (name->generation > 0) {
strbuf_grow(&sb, len +
1 + decimal_width(name->generation) +
1 + decimal_width(parent_number));
strbuf_addf(&sb, "%.*s~%d^%d", (int)len, name->tip_name,
name->generation, parent_number);
} else {
strbuf_grow(&sb, len +
1 + decimal_width(parent_number));
strbuf_addf(&sb, "%.*s^%d", (int)len, name->tip_name,
parent_number);
}
return strbuf_detach(&sb, NULL);
}
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
static void name_rev(struct commit *start_commit,
const char *tip_name, timestamp_t taggerdate,
int from_tag, int deref)
{
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
struct prio_queue queue;
struct commit *commit;
struct commit **parents_to_queue = NULL;
size_t parents_to_queue_nr, parents_to_queue_alloc = 0;
struct rev_name *start_name;
parse_commit(start_commit);
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
if (commit_is_before_cutoff(start_commit))
return;
start_name = create_or_update_name(start_commit, taggerdate, 0, 0,
from_tag);
if (!start_name)
return;
if (deref)
start_name->tip_name = xstrfmt("%s^0", tip_name);
else
start_name->tip_name = xstrdup(tip_name);
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
memset(&queue, 0, sizeof(queue)); /* Use the prio_queue as LIFO */
prio_queue_put(&queue, start_commit);
while ((commit = prio_queue_get(&queue))) {
struct rev_name *name = get_commit_rev_name(commit);
struct commit_list *parents;
int parent_number = 1;
parents_to_queue_nr = 0;
for (parents = commit->parents;
parents;
parents = parents->next, parent_number++) {
struct commit *parent = parents->item;
struct rev_name *parent_name;
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
int generation, distance;
parse_commit(parent);
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
if (commit_is_before_cutoff(parent))
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
continue;
if (parent_number > 1) {
generation = 0;
distance = name->distance + MERGE_TRAVERSAL_WEIGHT;
} else {
generation = name->generation + 1;
distance = name->distance + 1;
}
parent_name = create_or_update_name(parent, taggerdate,
generation,
distance, from_tag);
if (parent_name) {
if (parent_number > 1)
parent_name->tip_name =
get_parent_name(name,
parent_number);
else
parent_name->tip_name = name->tip_name;
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
ALLOC_GROW(parents_to_queue,
parents_to_queue_nr + 1,
parents_to_queue_alloc);
parents_to_queue[parents_to_queue_nr] = parent;
parents_to_queue_nr++;
}
}
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
/* The first parent must come out first from the prio_queue */
while (parents_to_queue_nr)
prio_queue_put(&queue,
parents_to_queue[--parents_to_queue_nr]);
}
name-rev: eliminate recursion in name_rev() The name_rev() function calls itself recursively for each interesting parent of the commit it got as parameter, and, consequently, it can segfault when processing a deep history if it exhausts the available stack space. E.g. running 'git name-rev --all' and 'git name-rev HEAD~100000' in the gcc, gecko-dev, llvm, and WebKit repositories results in segfaults on my machine ('ulimit -s' reports 8192kB of stack size limit), and nowadays the former segfaults in the Linux repo as well (it reached the necessasry depth sometime between v5.3-rc4 and -rc5). Eliminate the recursion by inserting the interesting parents into a LIFO 'prio_queue' [1] and iterating until the queue becomes empty. Note that the parent commits must be added in reverse order to the LIFO 'prio_queue', so their relative order is preserved during processing, i.e. the first parent should come out first from the queue, because otherwise performance greatly suffers on mergy histories [2]. The stacksize-limited test 'name-rev works in a deep repo' in 't6120-describe.sh' demonstrated this issue and expected failure. Now the recursion is gone, so flip it to expect success. Also gone are the dmesg entries logging the segfault of that segfaulting 'git name-rev' process on every execution of the test suite. Note that this slightly changes the order of lines in the output of 'git name-rev --all', usually swapping two lines every 35 lines in git.git or every 150 lines in linux.git. This shouldn't matter in practice, because the output has always been unordered anyway. This patch is best viewed with '--ignore-all-space'. [1] Early versions of this patch used a 'commit_list', resulting in ~15% performance penalty for 'git name-rev --all' in 'linux.git', presumably because of the memory allocation and release for each insertion and removal. Using a LIFO 'prio_queue' has basically no effect on performance. [2] We prefer shorter names, i.e. 'v0.1~234' is preferred over 'v0.1^2~5', meaning that usually following the first parent of a merge results in the best name for its ancestors. So when later we follow the remaining parent(s) of a merge, and reach an already named commit, then we usually find that we can't give that commit a better name, and thus we don't have to visit any of its ancestors again. OTOH, if we were to follow the Nth parent of the merge first, then the name of all its ancestors would include a corresponding '^N'. Those are not the best names for those commits, so when later we reach an already named commit following the first parent of that merge, then we would have to update the name of that commit and the names of all of its ancestors as well. Consequently, we would have to visit many commits several times, resulting in a significant slowdown. Signed-off-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2019-12-09 11:52:57 +00:00
clear_prio_queue(&queue);
free(parents_to_queue);
}
static int subpath_matches(const char *path, const char *filter)
{
const char *subpath = path;
while (subpath) {
if (!wildmatch(filter, subpath, 0))
return subpath - path;
subpath = strchr(subpath, '/');
if (subpath)
subpath++;
}
return -1;
}
static const char *name_ref_abbrev(const char *refname, int shorten_unambiguous)
{
if (shorten_unambiguous)
refname = shorten_unambiguous_ref(refname, 0);
else if (skip_prefix(refname, "refs/heads/", &refname))
; /* refname already advanced */
else
skip_prefix(refname, "refs/", &refname);
return refname;
}
struct name_ref_data {
int tags_only;
int name_only;
struct string_list ref_filters;
struct string_list exclude_filters;
};
static struct tip_table {
struct tip_table_entry {
struct object_id oid;
const char *refname;
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
struct commit *commit;
timestamp_t taggerdate;
unsigned int from_tag:1;
unsigned int deref:1;
} *table;
int nr;
int alloc;
int sorted;
} tip_table;
static void add_to_tip_table(const struct object_id *oid, const char *refname,
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
int shorten_unambiguous, struct commit *commit,
timestamp_t taggerdate, int from_tag, int deref)
{
refname = name_ref_abbrev(refname, shorten_unambiguous);
ALLOC_GROW(tip_table.table, tip_table.nr + 1, tip_table.alloc);
oidcpy(&tip_table.table[tip_table.nr].oid, oid);
tip_table.table[tip_table.nr].refname = xstrdup(refname);
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
tip_table.table[tip_table.nr].commit = commit;
tip_table.table[tip_table.nr].taggerdate = taggerdate;
tip_table.table[tip_table.nr].from_tag = from_tag;
tip_table.table[tip_table.nr].deref = deref;
tip_table.nr++;
tip_table.sorted = 0;
}
static int tipcmp(const void *a_, const void *b_)
{
const struct tip_table_entry *a = a_, *b = b_;
return oidcmp(&a->oid, &b->oid);
}
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
static int cmp_by_tag_and_age(const void *a_, const void *b_)
{
const struct tip_table_entry *a = a_, *b = b_;
int cmp;
/* Prefer tags. */
cmp = b->from_tag - a->from_tag;
if (cmp)
return cmp;
/* Older is better. */
if (a->taggerdate < b->taggerdate)
return -1;
return a->taggerdate != b->taggerdate;
}
static int name_ref(const char *path, const struct object_id *oid,
int flags UNUSED, void *cb_data)
{
struct object *o = parse_object(the_repository, oid);
struct name_ref_data *data = cb_data;
int can_abbreviate_output = data->tags_only && data->name_only;
int deref = 0;
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
int from_tag = 0;
struct commit *commit = NULL;
timestamp_t taggerdate = TIME_MAX;
if (data->tags_only && !starts_with(path, "refs/tags/"))
return 0;
if (data->exclude_filters.nr) {
struct string_list_item *item;
for_each_string_list_item(item, &data->exclude_filters) {
if (subpath_matches(path, item->string) >= 0)
return 0;
}
}
if (data->ref_filters.nr) {
struct string_list_item *item;
int matched = 0;
/* See if any of the patterns match. */
for_each_string_list_item(item, &data->ref_filters) {
/*
* Check all patterns even after finding a match, so
* that we can see if a match with a subpath exists.
* When a user asked for 'refs/tags/v*' and 'v1.*',
* both of which match, the user is showing her
* willingness to accept a shortened output by having
* the 'v1.*' in the acceptable refnames, so we
* shouldn't stop when seeing 'refs/tags/v1.4' matches
* 'refs/tags/v*'. We should show it as 'v1.4'.
*/
switch (subpath_matches(path, item->string)) {
case -1: /* did not match */
break;
case 0: /* matched fully */
matched = 1;
break;
default: /* matched subpath */
matched = 1;
can_abbreviate_output = 1;
break;
}
}
/* If none of the patterns matched, stop now */
if (!matched)
return 0;
}
while (o && o->type == OBJ_TAG) {
struct tag *t = (struct tag *) o;
if (!t->tagged)
break; /* broken repository */
o = parse_object(the_repository, &t->tagged->oid);
deref = 1;
taggerdate = t->date;
}
if (o && o->type == OBJ_COMMIT) {
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
commit = (struct commit *)o;
from_tag = starts_with(path, "refs/tags/");
if (taggerdate == TIME_MAX)
taggerdate = commit->date;
}
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
add_to_tip_table(oid, path, can_abbreviate_output, commit, taggerdate,
from_tag, deref);
return 0;
}
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
static void name_tips(void)
{
int i;
/*
* Try to set better names first, so that worse ones spread
* less.
*/
QSORT(tip_table.table, tip_table.nr, cmp_by_tag_and_age);
for (i = 0; i < tip_table.nr; i++) {
struct tip_table_entry *e = &tip_table.table[i];
if (e->commit) {
name_rev(e->commit, e->refname, e->taggerdate,
e->from_tag, e->deref);
}
}
}
static const struct object_id *nth_tip_table_ent(size_t ix, const void *table_)
{
const struct tip_table_entry *table = table_;
return &table[ix].oid;
}
static const char *get_exact_ref_match(const struct object *o)
{
int found;
if (!tip_table.table || !tip_table.nr)
return NULL;
if (!tip_table.sorted) {
QSORT(tip_table.table, tip_table.nr, tipcmp);
tip_table.sorted = 1;
}
found = oid_pos(&o->oid, tip_table.table, tip_table.nr,
nth_tip_table_ent);
if (0 <= found)
return tip_table.table[found].refname;
return NULL;
}
/* may return a constant string or use "buf" as scratch space */
static const char *get_rev_name(const struct object *o, struct strbuf *buf)
{
struct rev_name *n;
const struct commit *c;
if (o->type != OBJ_COMMIT)
return get_exact_ref_match(o);
c = (const struct commit *) o;
n = get_commit_rev_name(c);
if (!n)
return NULL;
if (!n->generation)
return n->tip_name;
else {
strbuf_reset(buf);
strbuf_addstr(buf, n->tip_name);
strbuf_strip_suffix(buf, "^0");
strbuf_addf(buf, "~%d", n->generation);
return buf->buf;
}
}
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 00:42:35 +00:00
static void show_name(const struct object *obj,
const char *caller_name,
int always, int allow_undefined, int name_only)
{
const char *name;
const struct object_id *oid = &obj->oid;
struct strbuf buf = STRBUF_INIT;
if (!name_only)
printf("%s ", caller_name ? caller_name : oid_to_hex(oid));
name = get_rev_name(obj, &buf);
if (name)
printf("%s\n", name);
else if (allow_undefined)
printf("undefined\n");
else if (always)
printf("%s\n", find_unique_abbrev(oid, DEFAULT_ABBREV));
else
die("cannot describe '%s'", oid_to_hex(oid));
strbuf_release(&buf);
}
static char const * const name_rev_usage[] = {
N_("git name-rev [<options>] <commit>..."),
N_("git name-rev [<options>] --all"),
N_("git name-rev [<options>] --annotate-stdin"),
NULL
};
static void name_rev_line(char *p, struct name_ref_data *data)
{
struct strbuf buf = STRBUF_INIT;
int counter = 0;
char *p_start;
const unsigned hexsz = the_hash_algo->hexsz;
for (p_start = p; *p; p++) {
#define ishex(x) (isdigit((x)) || ((x) >= 'a' && (x) <= 'f'))
if (!ishex(*p))
counter = 0;
else if (++counter == hexsz &&
!ishex(*(p+1))) {
struct object_id oid;
const char *name = NULL;
char c = *(p+1);
int p_len = p - p_start + 1;
counter = 0;
*(p+1) = 0;
if (!get_oid(p - (hexsz - 1), &oid)) {
struct object *o =
lookup_object(the_repository, &oid);
if (o)
name = get_rev_name(o, &buf);
}
*(p+1) = c;
if (!name)
continue;
if (data->name_only)
printf("%.*s%s", p_len - hexsz, p_start, name);
else
printf("%.*s (%s)", p_len, p_start, name);
p_start = p + 1;
}
}
/* flush */
if (p_start != p)
fwrite(p_start, p - p_start, 1, stdout);
strbuf_release(&buf);
}
int cmd_name_rev(int argc, const char **argv, const char *prefix)
{
struct object_array revs = OBJECT_ARRAY_INIT;
int all = 0, annotate_stdin = 0, transform_stdin = 0, allow_undefined = 1, always = 0, peel_tag = 0;
struct name_ref_data data = { 0, 0, STRING_LIST_INIT_NODUP, STRING_LIST_INIT_NODUP };
struct option opts[] = {
OPT_BOOL(0, "name-only", &data.name_only, N_("print only ref-based names (no object names)")),
OPT_BOOL(0, "tags", &data.tags_only, N_("only use tags to name the commits")),
OPT_STRING_LIST(0, "refs", &data.ref_filters, N_("pattern"),
N_("only use refs matching <pattern>")),
OPT_STRING_LIST(0, "exclude", &data.exclude_filters, N_("pattern"),
N_("ignore refs matching <pattern>")),
OPT_GROUP(""),
OPT_BOOL(0, "all", &all, N_("list all commits reachable from all refs")),
OPT_BOOL(0, "stdin", &transform_stdin, N_("deprecated: use --annotate-stdin instead")),
OPT_BOOL(0, "annotate-stdin", &annotate_stdin, N_("annotate text from stdin")),
OPT_BOOL(0, "undefined", &allow_undefined, N_("allow to print `undefined` names (default)")),
OPT_BOOL(0, "always", &always,
N_("show abbreviated commit object as fallback")),
{
/* A Hidden OPT_BOOL */
OPTION_SET_INT, 0, "peel-tag", &peel_tag, NULL,
N_("dereference tags in the input (internal use)"),
PARSE_OPT_NOARG | PARSE_OPT_HIDDEN, NULL, 1,
},
OPT_END(),
};
init_commit_rev_name(&rev_names);
git_config(git_default_config, NULL);
argc = parse_options(argc, argv, prefix, opts, name_rev_usage, 0);
if (transform_stdin) {
warning("--stdin is deprecated. Please use --annotate-stdin instead, "
"which is functionally equivalent.\n"
"This option will be removed in a future release.");
annotate_stdin = 1;
}
if (all + annotate_stdin + !!argc > 1) {
error("Specify either a list, or --all, not both!");
usage_with_options(name_rev_usage, opts);
}
if (all || annotate_stdin)
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
disable_cutoff();
for (; argc; argc--, argv++) {
struct object_id oid;
struct object *object;
struct commit *commit;
if (get_oid(*argv, &oid)) {
fprintf(stderr, "Could not get sha1 for %s. Skipping.\n",
*argv);
continue;
}
commit = NULL;
object = parse_object(the_repository, &oid);
if (object) {
struct object *peeled = deref_tag(the_repository,
object, *argv, 0);
if (peeled && peeled->type == OBJ_COMMIT)
commit = (struct commit *)peeled;
}
if (!object) {
fprintf(stderr, "Could not get object for %s. Skipping.\n",
*argv);
continue;
}
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
if (commit)
set_commit_cutoff(commit);
if (peel_tag) {
if (!commit) {
fprintf(stderr, "Could not get commit for %s. Skipping.\n",
*argv);
continue;
}
object = (struct object *)commit;
}
add_object_array(object, *argv, &revs);
}
name-rev: use generation numbers if available If a commit in a sequence of linear history has a non-monotonically increasing commit timestamp, git name-rev might not properly name the commit. This occurs because name-rev uses a heuristic of the commit date to avoid searching down tags which lead to commits that are older than the named commit. This is intended to avoid work on larger repositories. This heuristic impacts git name-rev, and by extension git describe --contains which is built on top of name-rev. Further more, if --all or --annotate-stdin is used, the heuristic is not enabled because the full history has to be analyzed anyways. This results in some confusion if a user sees that --annotate-stdin works but a normal name-rev does not. If the repository has a commit graph, we can use the generation numbers instead of using the commit dates. This is essentially the same check except that generation numbers make it exact, where the commit date heuristic could be incorrect due to clock errors. Since we're extending the notion of cutoff to more than one variable, create a series of functions for setting and checking the cutoff. This avoids duplication and moves access of the global cutoff and generation_cutoff to as few functions as possible. Add several test cases including a test that covers the new commitGraph behavior, as well as tests for --all and --annotate-stdin with and without commitGraphs. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-03-12 00:00:15 +00:00
adjust_cutoff_timestamp_for_slop();
for_each_ref(name_ref, &data);
name-rev: sort tip names before applying name_ref() is called for each ref and checks if its a better name for the referenced commit. If that's the case it remembers it and checks if a name based on it is better for its ancestors as well. This in done in the the order for_each_ref() imposes on us. That might not be optimal. If bad names happen to be encountered first (as defined by is_better_name()), names derived from them may spread to a lot of commits, only to be replaced by better names later. Setting better names first can avoid that. is_better_name() prefers tags, short distances and old references. The distance is a measure that we need to calculate for each candidate commit, but the other two properties are not dependent on the relationships of commits. Sorting the refs by them should yield better performance than the essentially random order we currently use. And applying older references first should also help to reduce rework due to the fact that older commits have less ancestors than newer ones. So add all details of names to the tip table first, then sort them to prefer tags and older references and then apply them in this order. Here's the performance as measures by hyperfine for the Linux repo before: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 851.1 ms ± 4.5 ms [User: 806.7 ms, System: 44.4 ms] Range (min … max): 845.9 ms … 859.5 ms 10 runs ... and with this patch: Benchmark #1: ./git -C ../linux/ name-rev --all Time (mean ± σ): 736.2 ms ± 8.7 ms [User: 688.4 ms, System: 47.5 ms] Range (min … max): 726.0 ms … 755.2 ms 10 runs Signed-off-by: René Scharfe <l.s.r@web.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-02-05 17:50:23 +00:00
name_tips();
if (annotate_stdin) {
struct strbuf sb = STRBUF_INIT;
while (strbuf_getline(&sb, stdin) != EOF) {
strbuf_addch(&sb, '\n');
name_rev_line(sb.buf, &data);
}
strbuf_release(&sb);
} else if (all) {
int i, max;
max = get_max_object_index();
for (i = 0; i < max; i++) {
struct object *obj = get_indexed_object(i);
if (!obj || obj->type != OBJ_COMMIT)
continue;
show_name(obj, NULL,
always, allow_undefined, data.name_only);
}
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 00:42:35 +00:00
} else {
int i;
for (i = 0; i < revs.nr; i++)
show_name(revs.objects[i].item, revs.objects[i].name,
always, allow_undefined, data.name_only);
Add "named object array" concept We've had this notion of a "object_list" for a long time, which eventually grew a "name" member because some users (notably git-rev-list) wanted to name each object as it is generated. That object_list is great for some things, but it isn't all that wonderful for others, and the "name" member is generally not used by everybody. This patch splits the users of the object_list array up into two: the traditional list users, who want the list-like format, and who don't actually use or want the name. And another class of users that really used the list as an extensible array, and generally wanted to name the objects. The patch is fairly straightforward, but it's also biggish. Most of it really just cleans things up: switching the revision parsing and listing over to the array makes things like the builtin-diff usage much simpler (we now see exactly how many members the array has, and we don't get the objects reversed from the order they were on the command line). One of the main reasons for doing this at all is that the malloc overhead of the simple object list was actually pretty high, and the array is just a lot denser. So this patch brings down memory usage by git-rev-list by just under 3% (on top of all the other memory use optimizations) on the mozilla archive. It does add more lines than it removes, and more importantly, it adds a whole new infrastructure for maintaining lists of objects, but on the other hand, the new dynamic array code is pretty obvious. The change to builtin-diff-tree.c shows a fairly good example of why an array interface is sometimes more natural, and just much simpler for everybody. Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
2006-06-20 00:42:35 +00:00
}
UNLEAK(revs);
return 0;
}