Currently, when looking for a packed object from the pack idx, a
simple binary search is used.
A conventional binary search loop looks like this:
unsigned lo, hi;
do {
unsigned mi = (lo + hi) / 2;
int cmp = "entry pointed at by mi" minus "target";
if (!cmp)
return mi; "mi is the wanted one"
if (cmp > 0)
hi = mi; "mi is larger than target"
else
lo = mi+1; "mi is smaller than target"
} while (lo < hi);
"did not find what we wanted"
The invariants are:
- When entering the loop, 'lo' points at a slot that is never
above the target (it could be at the target), 'hi' points at
a slot that is guaranteed to be above the target (it can
never be at the target).
- We find a point 'mi' between 'lo' and 'hi' ('mi' could be
the same as 'lo', but never can be as high as 'hi'), and
check if 'mi' hits the target. There are three cases:
- if it is a hit, we have found what we are looking for;
- if it is strictly higher than the target, we set it to
'hi', and repeat the search.
- if it is strictly lower than the target, we update 'lo'
to one slot after it, because we allow 'lo' to be at the
target and 'mi' is known to be below the target.
If the loop exits, there is no matching entry.
When choosing 'mi', we do not have to take the "middle" but
anywhere in between 'lo' and 'hi', as long as lo <= mi < hi is
satisfied. When we somehow know that the distance between the
target and 'lo' is much shorter than the target and 'hi', we
could pick 'mi' that is much closer to 'lo' than (hi+lo)/2,
which a conventional binary search would pick.
This patch takes advantage of the fact that the SHA-1 is a good
hash function, and as long as there are enough entries in the
table, we can expect uniform distribution. An entry that begins
with for example "deadbeef..." is much likely to appear much
later than in the midway of a reasonably populated table. In
fact, it can be expected to be near 87% (222/256) from the top
of the table.
This is a work-in-progress and has switches to allow easier
experiments and debugging. Exporting GIT_USE_LOOKUP environment
variable enables this code.
On my admittedly memory starved machine, with a partial KDE
repository (3.0G pack with 95M idx):
$ GIT_USE_LOOKUP=t git log -800 --stat HEAD >/dev/null
3.93user 0.16system 0:04.09elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+55588minor)pagefaults 0swaps
Without the patch, the numbers are:
$ git log -800 --stat HEAD >/dev/null
4.00user 0.15system 0:04.17elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+60258minor)pagefaults 0swaps
In the same repository:
$ GIT_USE_LOOKUP=t git log -2000 HEAD >/dev/null
0.12user 0.00system 0:00.12elapsed 97%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+4241minor)pagefaults 0swaps
Without the patch, the numbers are:
$ git log -2000 HEAD >/dev/null
0.05user 0.01system 0:00.07elapsed 100%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+8506minor)pagefaults 0swaps
There isn't much time difference, but the number of minor faults
seems to show that we are touching much smaller number of pages,
which is expected.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* mk/unpack-careful:
t5300: add test for "index-pack --strict"
receive-pack: allow using --strict mode for unpacking objects
unpack-objects: fix --strict handling
t5300: add test for "unpack-objects --strict"
unpack-objects: prevent writing of inconsistent objects
* fl/send-email-outside:
send-email: Don't require to be called in a repository
Git.pm: Don't require repository instance for ident
Git.pm: Don't require a repository instance for config
var: Don't require to be in a git repository.
When using git-svn to follow only a single (empty) path per
svn-remote (i.e. not using --stdlayout), following the history
of a renamed path was broken in
c586879cdf.
This reverts the regression for the single (emtpy) path per
svn-remote case.
To avoid breaking the tests in a committed revision, this is an
addendum to a patch originally submitted by
Santhosh Kumar Mani <santhoshmani@gmail.com>:
> git-svn: add test for renamed directory fetch
>
> This test tries to fetch a directory which had renames in the
> history from a SVN repository.
[ew: unneccesary dependency on the starting an HTTP server
removed from Santhosh's original test.]
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In commit 15387e3 (Test suite: reset TERM to its previous value after
testing., 2007-10-26), I added a workaround to reset TERM to its previous
value before the "test_done" at the end of "t7005-editor.sh" because
otherwise "test_done" would have printed the test result with a bad TERM
env variable (this resulted in output with no color on konsole).
But since commit c2116a1 (test-lib: fix TERM to dumb for test
repeatability, 2008-03-06), colored output is printed in a subshell with
TERM reset to its original value so the earlier workaround is not needed
anymore.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An earlier commit 4be6096 (apply --unidiff-zero: loosen sanity checks for
--unidiff=0 patches, 2006-09-17) made match_beginning and match_end
computed incorrectly. If a hunk inserts at the beginning, old position
recorded at the hunk is line 0, and if a hunk changes at the beginning, it
is line 1. The new test added to t4104 exposes that the old code did not
insist on matching at the beginning for a patch to add a line to an empty
file.
An even older 65aadb9 (apply: force matching at the beginning.,
2006-05-24) was equally wrong in that it tried to take hints from the
number of leading context lines, to decide if the hunk must match at the
beginning, but we can just look at the line number in the hunk to decide.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Peter Eriksen <s022018@student.dtu.dk>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With tcl/tk8.5 the lset command seems to behave differently. When
changing the background color through Edit->Preferences, the changes
are applied, but new dialogs, such as View->New view... barf with
Error: unknown color name "{#ffffff}"
Additionally when closing gitk, and starting it up again, a bad value
has been saved to ~/.gitk, preventing gitk from running properly; it
fails with
Error in startup script: unknown color name "{#ffffff}"
...
This commit fixes the problem by changing the color dialogs to pass
the empty string {} as the list index to choosecolor. This causes
the lset and lindex commands used by choosecolor to use and set the
whole variable (bgcolor, fgcolor or selectbgcolor) rather than
treating them as a 1-element list. Tested with tcl/tk8.4 and 8.5.
Dmitry Potapov reported this problem through
http://bugs.debian.org/472615
Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
It is a bit confusing on first read, that
"The packed archive format (.pack) is designed
to be unpackable..."
Signed-off-by: Peter Eriksen <s022018@student.dtu.dk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When git-fetch encounters the refspec "tag" it assumes that the next
argument will be a tag name. If there is no next argument, it should
die gracefully instead of erroring.
Signed-off-by: Kevin Ballard <kevin@sb.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This reverts commit 6aa6f92fda.
It caused is_deleted() subroutine to output warnings when dealing with
old, legacy gitweb blobdiff URLs without either 'hb' or 'hpb'
parameters.
This fixes http://bugs.debian.org/469083
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The earlier one did not correctly propagate GITWEB_CONFIG_SYSTEM from
Makefile to generated gitweb.cgi script.
Signed-off-by: Gerrit Pape <pape@smarden.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://repo.or.cz/git-gui:
git-gui: use +/- instead of ]/[ to show more/less context in diff
git-gui: Update french translation
git-gui: Switch keybindings for [ and ] to bracketleft and bracketright
On some systems, brackets cannot be used as event details
(they don't have a keysym), so use +/- instead (both on
keyboard and keypad) and add ctrl-= as a synonym of ctrl-+
for convenience.
[sp: Had to change accelerator to show only "$M1T-="; the
original version included "$M1T-+ $M1T-=" but this is
not drawn at all on Mac OS X.]
Signed-off-by: Michele Ballabio <barra_cuda@katamail.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
The rate of fixes that trickle in has slowed and we are definitely
getting there. Hopefully one final round and we will have the final
1.5.5 soon.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The interactive mode does not work with files whose names contain
characters that need C-quoting. `core.quotepath` configuration can be
used to work this limitation around to some degree, but backslash,
double-quote and control characters will still have problems.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* dd/cvsserver:
cvsserver: Use the user part of the email in log and annotate results
cvsserver: Add test for update -p
cvsserver: Implement update -p (print to stdout)
cvsserver: Add a few tests for 'status' command
cvsserver: Do not include status output for subdirectories if -l is passed
cvsserver: Only print the file part of the filename in status header
cvsserver: Respond to the 'editors' and 'watchers' commands
This test was already careful enough to skip signed tag tests if gpg
is not available, but it must also skip all verify tests, even those
that are about non-signed tags, because they also invoke gpg.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
af05d67 (Always set *nongit_ok in setup_git_directory_gently(),
2008-03-25) had a change from the patch originally submitted that resulted
in disabling aliases outside a git repository.
It turns out that some people used "alias.fubar = diff --color-words" in
$HOME/.gitconfig to use non-index diff (or any command that do not need
git repository) outside git repositories, and this change broke them,
so this resurrects the support for such usage.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Bound to Ctrl/Cmd + left & right square brackets, depending on
your platform.
[sp: Added missing binds for . to allow shortcuts to work when
not focused in the commit message area.]
Signed-off-by: Jonathan del Strother <jon.delStrother@bestbefore.tv>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Update the verify_tag() function to remove an unnecessary test, and add
additional check for angle brackets in the name and email field, and
spaces in the email field. The timestamp and timezone sections are made
more straight forward by using strspn().
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit d89c1df (filter-branch: don't use xargs -0, 2008-03-12) replaced a
'ls-files | xargs rm' pipeline by 'git clean'. 'git clean' however does
not recurse and remove directories by default.
Now, consider a tree-filter that renames a directory.
1. For the first commit everything works as expected
2. Then filter-branch checks out the files for the next commit. This
leaves the new directory behind because there is no real "branch
switching" involved that would notice that the directory can be
removed.
3. Then filter-branch invokes 'git clean' to remove exactly those
left-overs. But here it does not remove the directory.
4. The next tree-filter does not work as expected because there already
exists a directory with the new name.
Just add -d to 'git clean', so that empty directories are removed.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This test currently fails.
If b is a directory then 'mv a b' is not a plain "rename", but really a
"move", so we must also test that the directory does not exist with the
old name in the directory with the new name.
There's also some cleanup in the corresponding "rename file" test to avoid
spurious shell syntax errors and "ambigous ref" error from 'git show' (but
these should show up only if the test would fail anyway). Plus we also
test for the non-existence of the old file.
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at>
Since nearly its birth, git's tags have included a "tagger" field which
describes the name of tagger, email of tagger, and date and time of tagging.
But, this field was only loosely tested by git-mktag. Provide some thorough
testing for this field and also ensure that the tag header is separated
from the tag body by an empty line to reduce the convenience of creating
a flawed tag.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Earlier, if you changed a staged path into a directory in the work tree,
we happily ran lstat(2) on it and found that it exists, and declared that
the user changed it to a gitlink.
This is wrong for two reasons:
(1) It may be a directory, but it may not be a submodule, and in the
latter case, the change we need to report is "the blob at the path
has disappeared". We need to check with resolve_gitlink_ref() to be
consistent with what "git add" and "git update-index --add" does.
(2) lstat(2) may have succeeded only because a leading component of the
path was turned into a symbolic link that points at something that
exists in the work tree. In such a case, the path itself does not
exist anymore, as far as the index is concerned.
This fixes these breakages in diff-index that the previous patch has
exposed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff-index and diff-files can get confused in corner cases when an indexed
blob turns into something else in the work tree. This patch adds tests to
expose such breakages.
The test is classified under t2XXX series instead of t4XXX series, because
the ultimate objective is to fix "add -u" (and "commit -a" that shares the
same issue).
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Adds a gitcvs.dbtablenameprefix config variable, the contents of which
are prepended to any database tables names used by git-cvsserver. The
same substutions as gitcvs.dbname and gitcvs.dbuser are supported, and
any non-alphabetic characters are replaced with underscores.
A typo found in contrib/completion/git-completion.bash is also fixed.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Pass --quiet to cpio in git-clone to hide the (confusing) "0 blocks" message.
For compatibility with operating systems which might not support GNUisms,
the presence of --quiet is probed for by grepping cpio's --help output.
Signed-off-by: Bryan Donlan <bdonlan@fushizen.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Jonathan Scott Duff wrote:
> Recently I tried "git svn showignore" on my parrot repository and it
> failed. I tracked it down to the prop_walk() sub. When it recurses,
> $path has an extra / on the beginning (i.e., when it recurses, it
> tries to get the props for "//apps" instead of "/apps"). I *think*
> this is because $path is used in the recursive call rather than $p
> (which seems to contain a properly transformed $path). Anyway, I've
> attached a patch that works for me and I think is generally the right
> thing.
Patch-submitted-by: Jonathan Scott Duff
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
