Make use of the stream_loose_object() function introduced in the
preceding commit to unpack large objects. Before this we'd need to
malloc() the size of the blob before unpacking it, which could cause
OOM with very large blobs.
We could use the new streaming interface to unpack all blobs, but
doing so would be much slower, as demonstrated e.g. with this
benchmark using git-hyperfine[0]:
rm -rf /tmp/scalar.git &&
git clone --bare https://github.com/Microsoft/scalar.git /tmp/scalar.git &&
mv /tmp/scalar.git/objects/pack/*.pack /tmp/scalar.git/my.pack &&
git hyperfine \
-r 2 --warmup 1 \
-L rev origin/master,HEAD -L v "10,512,1k,1m" \
-s 'make' \
-p 'git init --bare dest.git' \
-c 'rm -rf dest.git' \
'./git -C dest.git -c core.bigFileThreshold={v} unpack-objects </tmp/scalar.git/my.pack'
Here we'll perform worse with lower core.bigFileThreshold settings
with this change in terms of speed, but we're getting lower memory use
in return:
Summary
'./git -C dest.git -c core.bigFileThreshold=10 unpack-objects </tmp/scalar.git/my.pack' in 'origin/master' ran
1.01 ± 0.01 times faster than './git -C dest.git -c core.bigFileThreshold=1k unpack-objects </tmp/scalar.git/my.pack' in 'origin/master'
1.01 ± 0.01 times faster than './git -C dest.git -c core.bigFileThreshold=1m unpack-objects </tmp/scalar.git/my.pack' in 'origin/master'
1.01 ± 0.02 times faster than './git -C dest.git -c core.bigFileThreshold=1m unpack-objects </tmp/scalar.git/my.pack' in 'HEAD'
1.02 ± 0.00 times faster than './git -C dest.git -c core.bigFileThreshold=512 unpack-objects </tmp/scalar.git/my.pack' in 'origin/master'
1.09 ± 0.01 times faster than './git -C dest.git -c core.bigFileThreshold=1k unpack-objects </tmp/scalar.git/my.pack' in 'HEAD'
1.10 ± 0.00 times faster than './git -C dest.git -c core.bigFileThreshold=512 unpack-objects </tmp/scalar.git/my.pack' in 'HEAD'
1.11 ± 0.00 times faster than './git -C dest.git -c core.bigFileThreshold=10 unpack-objects </tmp/scalar.git/my.pack' in 'HEAD'
A better benchmark to demonstrate the benefits of that this one, which
creates an artificial repo with a 1, 25, 50, 75 and 100MB blob:
rm -rf /tmp/repo &&
git init /tmp/repo &&
(
cd /tmp/repo &&
for i in 1 25 50 75 100
do
dd if=/dev/urandom of=blob.$i count=$(($i*1024)) bs=1024
done &&
git add blob.* &&
git commit -mblobs &&
git gc &&
PACK=$(echo .git/objects/pack/pack-*.pack) &&
cp "$PACK" my.pack
) &&
git hyperfine \
--show-output \
-L rev origin/master,HEAD -L v "512,50m,100m" \
-s 'make' \
-p 'git init --bare dest.git' \
-c 'rm -rf dest.git' \
'/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold={v} unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum'
Using this test we'll always use >100MB of memory on
origin/master (around ~105MB), but max out at e.g. ~55MB if we set
core.bigFileThreshold=50m.
The relevant "Maximum resident set size" lines were manually added
below the relevant benchmark:
'/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=50m unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'origin/master' ran
Maximum resident set size (kbytes): 107080
1.02 ± 0.78 times faster than '/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=512 unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'origin/master'
Maximum resident set size (kbytes): 106968
1.09 ± 0.79 times faster than '/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=100m unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'origin/master'
Maximum resident set size (kbytes): 107032
1.42 ± 1.07 times faster than '/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=100m unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'HEAD'
Maximum resident set size (kbytes): 107072
1.83 ± 1.02 times faster than '/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=50m unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'HEAD'
Maximum resident set size (kbytes): 55704
2.16 ± 1.19 times faster than '/usr/bin/time -v ./git -C dest.git -c core.bigFileThreshold=512 unpack-objects </tmp/repo/my.pack 2>&1 | grep Maximum' in 'HEAD'
Maximum resident set size (kbytes): 4564
This shows that if you have enough memory this new streaming method is
slower the lower you set the streaming threshold, but the benefit is
more bounded memory use.
An earlier version of this patch introduced a new
"core.bigFileStreamingThreshold" instead of re-using the existing
"core.bigFileThreshold" variable[1]. As noted in a detailed overview
of its users in [2] using it has several different meanings.
Still, we consider it good enough to simply re-use it. While it's
possible that someone might want to e.g. consider objects "small" for
the purposes of diffing but "big" for the purposes of writing them
such use-cases are probably too obscure to worry about. We can always
split up "core.bigFileThreshold" in the future if there's a need for
that.
0. https://github.com/avar/git-hyperfine/
1. https://lore.kernel.org/git/20211210103435.83656-1-chiyutianyi@gmail.com/
2. https://lore.kernel.org/git/20220120112114.47618-5-chiyutianyi@gmail.com/
Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Helped-by: Derrick Stolee <stolee@gmail.com>
Helped-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Han Xin <chiyutianyi@gmail.com>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As the name implies, "get_data(size)" will allocate and return a given
amount of memory. Allocating memory for a large blob object may cause the
system to run out of memory. Before preparing to replace calling of
"get_data()" to unpack large blob objects in latter commits, refactor
"get_data()" to reduce memory footprint for dry_run mode.
Because in dry_run mode, "get_data()" is only used to check the
integrity of data, and the returned buffer is not used at all, we can
allocate a smaller buffer and use it as zstream output. Make the function
return NULL in the dry-run mode, as no callers use the returned buffer.
The "find [...]objects/?? -type f | wc -l" test idiom being used here
is adapted from the same "find" use added to another test in
d9545c7f46 (fast-import: implement unpack limit, 2016-04-25).
Suggested-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Han Xin <chiyutianyi@gmail.com>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A mechanism to pack unreachable objects into a "cruft pack",
instead of ejecting them into loose form to be reclaimed later, has
been introduced.
* tb/cruft-packs:
sha1-file.c: don't freshen cruft packs
builtin/gc.c: conditionally avoid pruning objects via loose
builtin/repack.c: add cruft packs to MIDX during geometric repack
builtin/repack.c: use named flags for existing_packs
builtin/repack.c: allow configuring cruft pack generation
builtin/repack.c: support generating a cruft pack
builtin/pack-objects.c: --cruft with expiration
reachable: report precise timestamps from objects in cruft packs
reachable: add options to add_unseen_recent_objects_to_traversal
builtin/pack-objects.c: --cruft without expiration
builtin/pack-objects.c: return from create_object_entry()
t/helper: add 'pack-mtimes' test-tool
pack-mtimes: support writing pack .mtimes files
chunk-format.h: extract oid_version()
pack-write: pass 'struct packing_data' to 'stage_tmp_packfiles'
pack-mtimes: support reading .mtimes files
Documentation/technical: add cruft-packs.txt
A workflow change for translators are being proposed.
* jx/l10n-workflow-change:
l10n: Document the new l10n workflow
Makefile: add "po-init" rule to initialize po/XX.po
Makefile: add "po-update" rule to update po/XX.po
po/git.pot: don't check in result of "make pot"
po/git.pot: this is now a generated file
Makefile: remove duplicate and unwanted files in FOUND_SOURCE_FILES
i18n CI: stop allowing non-ASCII source messages in po/git.pot
Makefile: have "make pot" not "reset --hard"
Makefile: generate "po/git.pot" from stable LOCALIZED_C
Makefile: sort source files before feeding to xgettext
Teach "git repack --geometric" work better with "--keep-pack" and
avoid corrupting the repository when packsize limit is used.
* tb/geom-repack-with-keep-and-max:
builtin/repack.c: ensure that `names` is sorted
t7703: demonstrate object corruption with pack.packSizeLimit
repack: respect --keep-pack with geometric repack
"sparse-checkout" learns to work well with the sparse-index
feature.
* ds/sparse-sparse-checkout:
sparse-checkout: integrate with sparse index
p2000: add test for 'git sparse-checkout [add|set]'
sparse-index: complete partial expansion
sparse-index: partially expand directories
sparse-checkout: --no-sparse-index needs a full index
cache-tree: implement cache_tree_find_path()
sparse-index: introduce partially-sparse indexes
sparse-index: create expand_index()
t1092: stress test 'git sparse-checkout set'
t1092: refactor 'sparse-index contents' test
The multi-pack-index code did not protect the packfile it is going
to depend on from getting removed while in use, which has been
corrected.
* tb/midx-race-in-pack-objects:
builtin/pack-objects.c: ensure pack validity from MIDX bitmap objects
builtin/pack-objects.c: ensure included `--stdin-packs` exist
builtin/pack-objects.c: avoid redundant NULL check
pack-bitmap.c: check preferred pack validity when opening MIDX bitmap
Preliminary code refactoring around transport and bundle code.
* ds/bundle-uri:
bundle.h: make "fd" version of read_bundle_header() public
remote: allow relative_url() to return an absolute url
remote: move relative_url()
http: make http_get_file() external
fetch-pack: move --keep=* option filling to a function
fetch-pack: add a deref_without_lazy_fetch_extended()
dir API: add a generalized path_match_flags() function
connect.c: refactor sending of agent & object-format
Introduce a filesystem-dependent mechanism to optimize the way the
bits for many loose object files are ensured to hit the disk
platter.
* ns/batch-fsync:
core.fsyncmethod: performance tests for batch mode
t/perf: add iteration setup mechanism to perf-lib
core.fsyncmethod: tests for batch mode
test-lib-functions: add parsing helpers for ls-files and ls-tree
core.fsync: use batch mode and sync loose objects by default on Windows
unpack-objects: use the bulk-checkin infrastructure
update-index: use the bulk-checkin infrastructure
builtin/add: add ODB transaction around add_files_to_cache
cache-tree: use ODB transaction around writing a tree
core.fsyncmethod: batched disk flushes for loose-objects
bulk-checkin: rebrand plug/unplug APIs as 'odb transactions'
bulk-checkin: rename 'state' variable and separate 'plugged' boolean
Deprecate non-cone mode of the sparse-checkout feature.
* en/sparse-cone-becomes-default:
Documentation: some sparsity wording clarifications
git-sparse-checkout.txt: mark non-cone mode as deprecated
git-sparse-checkout.txt: flesh out pattern set sections a bit
git-sparse-checkout.txt: add a new EXAMPLES section
git-sparse-checkout.txt: shuffle some sections and mark as internal
git-sparse-checkout.txt: update docs for deprecation of 'init'
git-sparse-checkout.txt: wording updates for the cone mode default
sparse-checkout: make --cone the default
tests: stop assuming --no-cone is the default mode for sparse-checkout
"git add -i" was rewritten in C some time ago and has been in
testing; the reimplementation is now exposed to general public by
default.
* js/use-builtin-add-i:
add -i: default to the built-in implementation
t2016: require the PERL prereq only when necessary
Expose the new `git repack --cruft` mode from `git gc` via a new opt-in
flag. When invoked like `git gc --cruft`, `git gc` will avoid exploding
unreachable objects as loose ones, and instead create a cruft pack and
`.mtimes` file.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using cruft packs, the following race can occur when a geometric
repack that writes a MIDX bitmap takes place afterwords:
- First, create an unreachable object and do an all-into-one cruft
repack which stores that object in the repository's cruft pack.
- Then make that object reachable.
- Finally, do a geometric repack and write a MIDX bitmap.
Assuming that we are sufficiently unlucky as to select a commit from the
MIDX which reaches that object for bitmapping, then the `git
multi-pack-index` process will complain that that object is missing.
The reason is because we don't include cruft packs in the MIDX when
doing a geometric repack. Since the "make that object reachable" doesn't
necessarily mean that we'll create a new copy of that object in one of
the packs that will get rolled up as part of a geometric repack, it's
possible that the MIDX won't see any copies of that now-reachable
object.
Of course, it's desirable to avoid including cruft packs in the MIDX
because it causes the MIDX to store a bunch of objects which are likely
to get thrown away. But excluding that pack does open us up to the above
race.
This patch demonstrates the bug, and resolves it by including cruft
packs in the MIDX even when doing a geometric repack.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We use the `util` pointer for items in the `existing_packs` string list
to indicate which packs are going to be deleted. Since that has so far
been the only use of that `util` pointer, we just set it to 0 or 1.
But we're going to add an additional state to this field in the next
patch, so prepare for that by adding a #define for the first bit so we
can more expressively inspect the flags state.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In servers which set the pack.window configuration to a large value, we
can wind up spending quite a lot of time finding new bases when breaking
delta chains between reachable and unreachable objects while generating
a cruft pack.
Introduce a handful of `repack.cruft*` configuration variables to
control the parameters used by pack-objects when generating a cruft
pack.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Expose a way to split the contents of a repository into a main and cruft
pack when doing an all-into-one repack with `git repack --cruft -d`, and
a complementary configuration variable.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a previous patch, pack-objects learned how to generate a cruft pack
so long as no objects are dropped.
This patch teaches pack-objects to handle the case where a non-never
`--cruft-expiration` value is passed. This case is slightly more
complicated than before, because we want pack-objects to save
unreachable objects which would have been pruned when there is another
recent (i.e., non-prunable) unreachable object which reaches the other.
We'll call these objects "unreachable but reachable-from-recent".
Here is how pack-objects handles `--cruft-expiration`:
- Instead of adding all objects outside of the kept pack(s) into the
packing list, only handle the ones whose mtime is within the grace
period.
- Construct a reachability traversal whose tips are the
unreachable-but-recent objects.
- Then, walk along that traversal, stopping if we reach an object in
the kept pack. At each step along the traversal, we add the object
we are visiting to the packing list.
In the majority of these cases, any object we visit in this traversal
will already be in our packing list. But we will sometimes encounter
reachable-from-recent cruft objects, which we want to retain even if
they aged out of the grace period.
The most subtle point of this process is that we actually don't need to
bother to update the rescued object's mtime. Even though we will write
an .mtimes file with a value that is older than the expiration window,
it will continue to survive cruft repacks so long as any objects which
reach it haven't aged out.
That is, a future repack will also exclude that object from the initial
packing list, only to discover it later on when doing the reachability
traversal.
Finally, stopping early once an object is found in a kept pack is safe
to do because the kept packs ordinarily represent which packs will
survive after repacking. Assuming that it _isn't_ safe to halt a
traversal early would mean that there is some ancestor object which is
missing, which implies repository corruption (i.e., the complete set of
reachable objects isn't present).
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function behaves very similarly to what we will need in
pack-objects in order to implement cruft packs with expiration. But it
is lacking a couple of things. Namely, it needs:
- a mechanism to communicate the timestamps of individual recent
objects to some external caller
- and, in the case of packed objects, our future caller will also want
to know the originating pack, as well as the offset within that pack
at which the object can be found
- finally, it needs a way to skip over packs which are marked as kept
in-core.
To address the first two, add a callback interface in this patch which
reports the time of each recent object, as well as a (packed_git,
off_t) pair for packed objects.
Likewise, add a new option to the packed object iterators to skip over
packs which are marked as kept in core. This option will become
implicitly tested in a future patch.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach `pack-objects` how to generate a cruft pack when no objects are
dropped (i.e., `--cruft-expiration=never`). Later patches will teach
`pack-objects` how to generate a cruft pack that prunes objects.
When generating a cruft pack which does not prune objects, we want to
collect all unreachable objects into a single pack (noting and updating
their mtimes as we accumulate them). Ordinary use will pass the result
of a `git repack -A` as a kept pack, so when this patch says "kept
pack", readers should think "reachable objects".
Generating a non-expiring cruft packs works as follows:
- Callers provide a list of every pack they know about, and indicate
which packs are about to be removed.
- All packs which are going to be removed (we'll call these the
redundant ones) are marked as kept in-core.
Any packs the caller did not mention (but are known to the
`pack-objects` process) are also marked as kept in-core. Packs not
mentioned by the caller are assumed to be unknown to them, i.e.,
they entered the repository after the caller decided which packs
should be kept and which should be discarded.
Since we do not want to include objects in these "unknown" packs
(because we don't know which of their objects are or aren't
reachable), these are also marked as kept in-core.
- Then, we enumerate all objects in the repository, and add them to
our packing list if they do not appear in an in-core kept pack.
This results in a new cruft pack which contains all known objects that
aren't included in the kept packs. When the kept pack is the result of
`git repack -A`, the resulting pack contains all unreachable objects.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A new caller in the next commit will want to immediately modify the
object_entry structure created by create_object_entry(). Instead of
forcing that caller to wastefully look-up the entry we just created,
return it from create_object_entry() instead.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This structure will be used to communicate the per-object mtimes when
writing a cruft pack. Here, we need the full packing_data structure
because the mtime information is stored in an array there, not on the
individual object_entry's themselves (to avoid paying the overhead in
structure width for operations which do not generate a cruft pack).
We haven't passed this information down before because one of the two
callers (in bulk-checkin.c) does not have a packing_data structure at
all. In that case (where no cruft pack will be generated), NULL is
passed instead.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To store the individual mtimes of objects in a cruft pack, introduce a
new `.mtimes` format that can optionally accompany a single pack in the
repository.
The format is defined in Documentation/technical/pack-format.txt, and
stores a 4-byte network order timestamp for each object in name (index)
order.
This patch prepares for cruft packs by defining the `.mtimes` format,
and introducing a basic API that callers can use to read out individual
mtimes.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git remote -v" now shows the list-objects-filter used during
fetching from the remote, if available.
* ac/remote-v-with-object-list-filters:
builtin/remote.c: teach `-v` to list filters for promisor remotes
"git -c branch.autosetupmerge=simple branch $A $B" will set the $B
as $A's upstream only when $A and $B shares the same name, and "git
-c push.default=simple" on branch $A would push to update the
branch $A at the remote $B came from. Also more places use the
sole remote, if exists, before defaulting to 'origin'.
* tk/simple-autosetupmerge:
push: new config option "push.autoSetupRemote" supports "simple" push
push: default to single remote even when not named origin
branch: new autosetupmerge option 'simple' for matching branches
In the preceding commit we moved away from using xgettext(1) to both
generate the po/git.pot, and to merge the incrementally generated
po/git.pot+ file as we sourced translations from C, shell and Perl.
Doing it this way, which dates back to my initial
implementation[1][2][3] was conflating two things: With xgettext(1)
the --from-code both controls what encoding is specified in the
po/git.pot's header, and what encoding we allow in source messages.
We don't ever want to allow non-ASCII in *source messages*, and doing
so has hid e.g. a buggy message introduced in
a6226fd772 (submodule--helper: convert the bulk of cmd_add() to C,
2021-08-10) from us, we'd warn about it before, but only when running
"make pot", but the operation would still succeed. Now we'll error out
on it when running "make pot".
Since the preceding Makefile changes made this easy: let's add a "make
check-pot" target with the same prerequisites as the "po/git.pot"
target, but without changing the file "po/git.pot". Running it as part
of the "static-analysis" CI target will ensure that we catch any such
issues in the future. E.g.:
$ make check-pot
XGETTEXT .build/pot/po/builtin/submodule--helper.c.po
xgettext: Non-ASCII string at builtin/submodule--helper.c:3381.
Please specify the source encoding through --from-code.
make: *** [.build/pot/po/builtin/submodule--helper.c.po] Error 1
1. cd5513a716 (i18n: Makefile: "pot" target to extract messages
marked for translation, 2011-02-22)
2. adc3b2b276 (Makefile: add xgettext target for *.sh files,
2011-05-14)
3. 5e9637c629 (i18n: add infrastructure for translating Git with
gettext, 2011-11-18)
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Jiang Xin <zhiyou.jx@alibaba-inc.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git fetch --recurse-submodules" from multiple remotes (either from
a remote group, or "--all") used to make one extra "git fetch" in
the submodules, which has been corrected.
* jc/avoid-redundant-submodule-fetch:
fetch: do not run a redundant fetch from submodule
The way "git fetch" without "--update-head-ok" ensures that HEAD in
no worktree points at any ref being updated was too wasteful, which
has been optimized a bit.
* os/fetch-check-not-current-branch:
fetch: limit shared symref check only for local branches
The "--current" option of "git show-branch" should have been made
incompatible with the "--reflog" mode, but this was not enforced,
which has been corrected.
* jc/show-branch-g-current:
show-branch: -g and --current are incompatible
When using a multi-pack bitmap, pack-objects will try to perform its
traversal using a call to `traverse_bitmap_commit_list()`, which calls
`add_object_entry_from_bitmap()` to add each object it finds to its
packing list.
This path can cause pack-objects to add objects from packs that don't
have open pack_fds on them, by avoiding a call to `is_pack_valid()`.
This is because we only call `is_pack_valid()` on the preferred pack (in
order to do verbatim reuse via `reuse_partial_packfile_from_bitmap()`)
and not others when loading a MIDX bitmap.
In this case, `add_object_entry_from_bitmap()` will check whether it
wants each object entry by calling `want_object_in_pack()`, which will
call `want_found_object` (since its caller already supplied a
`found_pack`). In most cases (particularly without `--local`, and when
`ignored_packed_keep_on_disk` and `ignored_packed_keep_in_core` are
both "0"), we'll take the entry from the pack contained in the MIDX
bitmap, all without an open pack_fd.
When we then try to use that entry later to assemble the actual pack,
we'll be susceptible to any simultaneous writers moving that pack out of
the way (e.g., due to a concurrent repack) without having an open file
descriptor, causing races that result in errors like:
remote: Enumerating objects: 1498802, done.
remote: fatal: packfile ./objects/pack/pack-e57d433b5a588daa37fbe946e2b28dfaec03a93e.pack cannot be accessed
remote: aborting due to possible repository corruption on the remote side.
This race can happen even with multi-pack bitmaps, since we may open a
MIDX bitmap that is being rewritten long before its packs are actually
unlinked.
Work around this by calling `is_pack_valid()` from within
`want_found_object()`, matching the behavior in
`want_object_in_pack_one()` (which has an analogous call). Most calls to
`is_pack_valid()` should be basically no-ops, since only the first call
requires us to open a file (subsequent calls realize the file is already
open, and return immediately).
Importantly, when `want_object_in_pack()` is given a non-NULL
`*found_pack`, but `want_found_object()` rejects the copy of the object
in that pack, we must reset `*found_pack` and `*found_offset` to NULL
and 0, respectively. Failing to do so could lead to other checks in
`want_object_in_pack()` (such as `want_object_in_pack_one()`) using the
same (invalid) pack as `*found_pack`, meaning that we don't call
`is_pack_valid()` because `p == *found_pack`. This can lead the caller
to believe it can use a copy of an object from an invalid pack.
An alternative approach to closing this race would have been to call
`is_pack_valid()` on _all_ packs in a multi-pack bitmap on load. This
has a couple of problems:
- it is unnecessarily expensive in the cases where we don't actually
need to open any packs (e.g., in `git rev-list --use-bitmap-index
--count`)
- more importantly, it means any time we would have hit this race,
we'll avoid using bitmaps altogether, leading to significant
slowdowns by forcing a full object traversal
Co-authored-by: Victoria Dye <vdye@github.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A subsequent patch will teach `want_object_in_pack()` to set its
`*found_pack` and `*found_offset` poitners to NULL when the provided
pack does not pass the `is_pack_valid()` check.
The `--stdin-packs` mode of `pack-objects` is not quite prepared to
handle this. To prepare it for this change, do the following two things:
- Ensure provided packs pass the `is_pack_valid()` check when
collecting the caller-provided packs into the "included" and
"excluded" lists.
- Gracefully handle any _invalid_ packs being passed to
`want_object_in_pack()`.
Calling `is_pack_valid()` early on makes it substantially less likely
that we will have to deal with a pack going away, since we'll have an
open file descriptor on its contents much earlier.
But even packs with open descriptors can become invalid in the future if
we (a) hit our open descriptor limit, forcing us to close some open
packs, and (b) one of those just-closed packs has gone away in the
meantime.
`add_object_entry_from_pack()` depends on having a non-NULL
`*found_pack`, since it passes that pointer to `packed_object_info()`,
meaning that we would SEGV if the pointer became NULL (like we propose
to do in `want_object_in_pack()` in the following patch).
But avoiding calling `packed_object_info()` entirely is OK, too, since
its only purpose is to identify which objects in the included packs are
commits, so that they can form the tips of the advisory traversal used
to discover the object namehashes.
Failing to do this means that at worst we will produce lower-quality
deltas, but it does not prevent us from generating the pack as long as
we can find a copy of each object from the disappearing pack in some
other part of the repository.
Co-authored-by: Victoria Dye <vdye@github.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Before calling `for_each_object_in_pack()`, the caller
`read_packs_list_from_stdin()` loops through each of the `include_packs`
and checks that its `->util` pointer (which is used to store the `struct
packed_git *` itself) is non-NULL.
This check is redundant, because `read_packs_list_from_stdin()` already
checks that the included packs are non-NULL earlier on in the same
function (and it does not add any new entries in between).
Remove this check, since it is not doing anything in the meantime.
Co-authored-by: Victoria Dye <vdye@github.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When modifying the sparse-checkout definition, the sparse-checkout
builtin calls update_sparsity() to modify the SKIP_WORKTREE bits of all
cache entries in the index. Before, we needed the index to be fully
expanded in order to ensure we had the full list of files necessary that
match the new patterns.
Insert a call to reset_sparse_directories() that expands sparse
directories that are within the new pattern list, but only far enough
that every necessary file path now exists as a cache entry. The
remaining logic within update_sparsity() will modify the SKIP_WORKTREE
bits appropriately.
This allows us to disable command_requires_full_index within the
sparse-checkout builtin. Add tests that demonstrate that we are not
expanding to a full index unnecessarily.
We can see the improved performance in the p2000 test script:
Test HEAD~1 HEAD
------------------------------------------------------------------------
2000.24: git ... (sparse-v3) 2.14(1.55+0.58) 1.57(1.03+0.53) -26.6%
2000.25: git ... (sparse-v4) 2.20(1.62+0.57) 1.58(0.98+0.59) -28.2%
These reductions of 26-28% are small compared to most examples, but the
time is dominated by writing a new copy of the base repository to the
worktree and then deleting it again. The fact that the previous index
expansion was such a large portion of the time is telling how important
it is to complete this sparse index integration.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the --no-sparse-index option is supplied, the sparse-checkout
builtin should explicitly ask to expand a sparse index to a full one.
This is currently done implicitly due to the command_requires_full_index
protection, but that will be removed in an upcoming change.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A future change will present a temporary, in-memory mode where the index
can both contain sparse directory entries but also not be completely
collapsed to the smallest possible sparse directories. This will be
necessary for modifying the sparse-checkout definition while using a
sparse index.
For now, convert the single-bit member 'sparse_index' in 'struct
index_state' to be a an 'enum sparse_index_mode' with three modes:
* INDEX_EXPANDED (0): No sparse directories exist. This is always the
case for repositories that do not use cone-mode sparse-checkout.
* INDEX_COLLAPSED: Sparse directories may exist. Files outside the
sparse-checkout cone are reduced to sparse directory entries whenever
possible.
* INDEX_PARTIALLY_SPARSE: Sparse directories may exist. Some file
entries outside the sparse-checkout cone may exist. Running
convert_to_sparse() may further reduce those files to sparse directory
entries.
The main reason to store this extra information is to allow
convert_to_sparse() to short-circuit when the index is already in
INDEX_EXPANDED mode but to actually do the necessary work when in
INDEX_PARTIALLY_SPARSE mode.
The INDEX_PARTIALLY_SPARSE mode will be used in an upcoming change.
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce and apply coccinelle rule to discourage an explicit
comparison between a pointer and NULL, and applies the clean-up to
the maintenance track.
* ep/maint-equals-null-cocci:
tree-wide: apply equals-null.cocci
tree-wide: apply equals-null.cocci
contrib/coccinnelle: add equals-null.cocci
"git show :<path>" learned to work better with the sparse-index
feature.
* ds/sparse-colon-path:
rev-parse: integrate with sparse index
object-name: diagnose trees in index properly
object-name: reject trees found in the index
show: integrate with the sparse index
t1092: add compatibility tests for 'git show'
Teach "git stash" to work better with sparse index entries.
* vd/sparse-stash:
unpack-trees: preserve index sparsity
stash: apply stash using 'merge_ort_nonrecursive()'
read-cache: set sparsity when index is new
sparse-index: expose 'is_sparse_index_allowed()'
stash: integrate with sparse index
stash: expand sparse-checkout compatibility testing
"git bisect" was too silent before it is ready to start computing
the actual bisection, which has been corrected.
* cd/bisect-messages-from-pre-flight-states:
bisect: output bisect setup status in bisect log
bisect: output state before we are ready to compute bisection
"git pull" without "--recurse-submodules=<arg>" made
submodule.recurse take precedence over fetch.recurseSubmodules by
mistake, which has been corrected.
* gc/pull-recurse-submodules:
pull: do not let submodule.recurse override fetch.recurseSubmodules
The previous patch demonstrates a scenario where the list of packs
written by `pack-objects` (and stored in the `names` string_list) is
out-of-order, and can thus cause us to delete packs we shouldn't.
This patch resolves that bug by ensuring that `names` is sorted in all
cases, not just when
delete_redundant && pack_everything & ALL_INTO_ONE
is true.
Because we did sort `names` in that case (which, prior to `--geometric`
repacks, was the only time we would actually delete packs, this is only
a bug for `--geometric` repacks.
It would be sufficient to only sort `names` when `delete_redundant` is
set to a non-zero value. But sorting a small list of strings is cheap,
and it is defensive against future calls to `string_list_has_string()`
on this list.
Co-discovered-by: Victoria Dye <vdye@github.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update 'repack' to ignore packs named on the command line with the
'--keep-pack' option. Specifically, modify 'init_pack_geometry()' to treat
command line-kept packs the same way it treats packs with an on-disk '.keep'
file (that is, skip the pack and do not include it in the 'geometry'
structure).
Without this handling, a '--keep-pack' pack would be included in the
'geometry' structure. If the pack is *before* the geometry split line (with
at least one other pack and/or loose objects present), 'repack' assumes the
pack's contents are "rolled up" into another pack via 'pack-objects'.
However, because the internally-invoked 'pack-objects' properly excludes
'--keep-pack' objects, any new pack it creates will not contain the kept
objects. Finally, 'repack' deletes the '--keep-pack' as "redundant" (since
it assumes 'pack-objects' created a new pack with its contents), resulting
in possible object loss and repository corruption.
Add a test ensuring that '--keep-pack' packs are now appropriately handled.
Co-authored-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Victoria Dye <vdye@github.com>
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In c7c4bdeccf (run-command API: remove "env" member, always use
"env_array", 2021-11-25), there was a push to replace
cld.env = env->v;
with
strvec_pushv(&cld.env_array, env->v);
The conversion in c7c4bdeccf was mostly plug-and-play, with the snag
that some instances of strvec_pushv() became guarded with a NULL check
to ensure that the second argument was non-NULL.
This conversion was slightly over-eager to add a conditional in
builtin/receive-pack.c::unpack(), since we know at the point that we
add the result of `tmp_objdir_env()` into the child process's
environment, that `tmp_objdir` is non-NULL.
This follows from the conditional just before our strvec_pushv() call
(which returns from the function if `tmp_objdir` was NULL), as well as
the call to tmp_objdir_add_as_alternate() just below, which relies on
its argument (`tmp_objdir`) being non-NULL.
In the meantime, this extra conditional isn't hurting anything. But it
is redundant and thus unnecessarily confusing. So let's remove it.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When 7dce19d3 (fetch/pull: Add the --recurse-submodules option,
2010-11-12) introduced the "--recurse-submodule" option, the
approach taken was to perform fetches in submodules only once, after
all the main fetching (it may usually be a fetch from a single
remote, but it could be fetching from a group of remotes using
fetch_multiple()) succeeded. Later we added "--all" to fetch from
all defined remotes, which complicated things even more.
If your project has a submodule, and you try to run "git fetch
--recurse-submodule --all", you'd see a fetch for the top-level,
which invokes another fetch for the submodule, followed by another
fetch for the same submodule. All but the last fetch for the
submodule come from a "git fetch --recurse-submodules" subprocess
that is spawned via the fetch_multiple() interface for the remotes,
and the last fetch comes from the code at the end.
Because recursive fetching from submodules is done in each fetch for
the top-level in fetch_multiple(), the last fetch in the submodule
is redundant. It only matters when fetch_one() interacts with a
single remote at the top-level.
While we are at it, there is one optimization that exists in dealing
with a group of remote, but is missing when "--all" is used. In the
former, when the group turns out to be a group of one, instead of
spawning "git fetch" as a subprocess via the fetch_multiple()
interface, we use the normal fetch_one() code path. Do the same
when handing "--all", if it turns out that we have only one remote
defined.
Reviewed-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This method was initially written in 63e95beb0 (submodule: port
resolve_relative_url from shell to C, 2016-05-15). As we will need
similar functionality in the bundle URI feature, extract this to be
available in remote.h.
The code is almost exactly the same, except for the following trivial
differences:
* Fix whitespace and wrapping issues with the prototype and argument
lists.
* Let's call starts_with_dot_{,dot_}slash_native() instead of the
functionally identical "starts_with_dot_{,dot_}slash()" wrappers
"builtin/submodule--helper.c".
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a path_match_flags() function and have the two sets of
starts_with_dot_{,dot_}slash() functions added in
63e95beb08 (submodule: port resolve_relative_url from shell to C,
2016-04-15) and a2b26ffb1a (fsck: convert gitmodules url to URL
passed to curl, 2020-04-18) be thin wrappers for it.
As the latter of those notes the fsck version was copied from the
initial builtin/submodule--helper.c version.
Since the code added in a2b26ffb1a was doing really doing the same as
win32_is_dir_sep() added in 1cadad6f65 (git clone <url>
C:\cygwin\home\USER\repo' is working (again), 2018-12-15) let's move
the latter to git-compat-util.h is a is_xplatform_dir_sep(). We can
then call either it or the platform-specific is_dir_sep() from this
new function.
Let's likewise change code in various other places that was hardcoding
checks for "'/' || '\\'" with the new is_xplatform_dir_sep(). As can
be seen in those callers some of them still concern themselves with
':' (Mac OS classic?), but let's leave the question of whether that
should be consolidated for some other time.
As we expect to make wider use of the "native" case in the future,
define and use two starts_with_dot_{,dot_}slash_native() convenience
wrappers. This makes the diff in builtin/submodule--helper.c much
smaller.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Derrick Stolee <derrickstolee@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This check was introduced in 8ee5d73137 (Fix fetch/pull when run without
--update-head-ok, 2008-10-13) in order to protect against replacing the ref
of the active branch by mistake, for example by running git fetch origin
master:master.
It was later extended in 8bc1f39f41 (fetch: protect branches checked out
in all worktrees, 2021-12-01) to scan all worktrees.
This operation is very expensive (takes about 30s in my repository) when
there are many tags or branches, and it is executed on every fetch, even if
no local heads are updated at all.
Limit it to protect only refs/heads/* to improve fetch performance.
Signed-off-by: Orgad Shaneh <orgads@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 4c28e4ada0 (commit: die before asking to edit the log
message, 2010-12-20), we have been "leaking" the "author_ident" when
prepare_to_commit() fails. Instead of returning from right there,
introduce an exit status variable and jump to the clean-up label
at the end.
Instead of explicitly releasing the resource with strbuf_release(),
mark the variable with UNLEAK() at the end, together with two other
variables that are already marked as such. If this were in a
utility function that is called number of times, but these are
different, we should explicitly release resources that grow
proportionally to the size of the problem being solved, but
cmd_commit() is like main() and there is no point in spending extra
cycles to release individual pieces of resource at the end, just
before process exit will clean everything for us for free anyway.
This fixes a leak demonstrated by e.g. "t3505-cherry-pick-empty.sh",
but unfortunately we cannot mark it or other affected tests as passing
now with "TEST_PASSES_SANITIZE_LEAK=true" as we'll need to fix many
other memory leaks before doing so.
Incidentally there are two tests that always passes the leak checker
with or without this change. Mark them as such.
This is based on an earlier patch by Ævar, but takes a different
approach that is more maintainable.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
