run-command: treat inaccessible directories as ENOENT
When execvp reports EACCES, it can be one of two things:
1. We found a file to execute, but did not have
permissions to do so.
2. We did not have permissions to look in some directory
in the $PATH.
In the former case, we want to consider this a
permissions problem and report it to the user as such (since
getting this for something like "git foo" is likely a
configuration error).
In the latter case, there is a good chance that the
inaccessible directory does not contain anything of
interest. Reporting "permission denied" is confusing to the
user (and prevents our usual "did you mean...?" lookup). It
also prevents git from trying alias lookup, since we do so
only when an external command does not exist (not when it
exists but has an error).
This patch detects EACCES from execvp, checks whether we are
in case (2), and if so converts errno to ENOENT. This
behavior matches that of "bash" (but not of simpler shells
that use execvp more directly, like "dash").
Test stolen from Junio.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
compat/mingw.[ch]: Change return type of exec functions to int
The POSIX standard specifies a return type of int for all six exec
functions. In addition, all exec functions return -1 on error, and
simply do not return on success. However, the current emulation of
the exec functions on mingw are declared with a void return type.
This would cause a problem should any code attempt to call the
exec function in a non-void context. In particular, if an exec
function were used in a conditional it would fail to compile.
In order to improve the fidelity of the emulation, we change the
return type of the mingw_execv[p] functions to int and return -1
on error.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
push: error out when the "upstream" semantics does not make sense
The user can say "git push" without specifying any refspec. When using
the "upstream" semantics via the push.default configuration, the user
wants to update the "upstream" branch of the current branch, which is the
branch at a remote repository the current branch is set to integrate with,
with this command.
However, there are cases that such a "git push" that uses the "upstream"
semantics does not make sense:
- The current branch does not have branch.$name.remote configured. By
definition, "git push" that does not name where to push to will not
know where to push to. The user may explicitly say "git push $there",
but again, by definition, no branch at repository $there is set to
integrate with the current branch in this case and we wouldn't know
which remote branch to update.
- The current branch does have branch.$name.remote configured, but it
does not specify branch.$name.merge that names what branch at the
remote this branch integrates with. "git push" knows where to push in
this case (or the user may explicitly say "git push $remote" to tell us
where to push), but we do not know which remote branch to update.
- The current branch does have its remote and upstream branch configured,
but the user said "git push $there", where $there is not the remote
named by "branch.$name.remote". By definition, no branch at repository
$there is set to integrate with the current branch in this case, and
this push is not meant to update any branch at the remote repository
$there.
The first two cases were already checked correctly, but the third case was
not checked and we ended up updating the branch named branch.$name.merge
at repository $there, which was totally bogus.
add--interactive: ignore unmerged entries in patch mode
When "add -p" sees an unmerged entry, it shows the combined
diff and then immediately skips the hunk. This can be
confusing in a variety of ways, depending on whether there
are other changes to stage (in which case you get the
superfluous combined diff output in between other hunks) or
not (in which case you get the combined diff and the program
exits immediately, rather than seeing "No changes").
The current behavior was not planned, and is just what the
implementation happens to do. Instead, let's explicitly
remove unmerged entries from our list of modified files, and
print a warning that we are ignoring them.
We can cheaply find which entries are unmerged by adding
"--raw" output to the "diff-files --numstat" we already run.
There is one non-obvious thing we must change when parsing
this combined output. Before this patch, when we saw a
numstat line for a file that did not have index changes, we
would create a new record with 'unchanged' in the 'INDEX'
field. Because "--raw" comes before "--numstat", we must
move this special-case down to the raw-line case (and it is
sufficient to move it rather than handle it in both places,
since any file which has a --numstat will also have a --raw
entry).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-commit: rephrase the "initial-ness" of templates
The description of "commit -t <file>" said the file is used "as the
initial version" of the commit message, but in the context of an SCM,
"version" is a loaded word that can needlesslyl confuse readers.
Explain the purpose of the mechanism without using "version".
fetch-pack: new --stdin option to read refs from stdin
If a remote repo has too many tags (or branches), cloning it over the
smart HTTP transport can fail because remote-curl.c puts all the refs
from the remote repo on the fetch-pack command line. This can make the
command line longer than the global OS command line limit, causing
fetch-pack to fail.
This is especially a problem on Windows where the command line limit is
orders of magnitude shorter than Linux. There are already real repos out
there that msysGit cannot clone over smart HTTP due to this problem.
Here is an easy way to trigger this problem:
git init too-many-refs
cd too-many-refs
echo bla > bla.txt
git add .
git commit -m test
sha=$(git rev-parse HEAD)
tag=$(perl -e 'print "bla" x 30')
for i in `seq 50000`; do
echo $sha refs/tags/$tag-$i >> .git/packed-refs
done
Then share this repo over the smart HTTP protocol and try cloning it:
$ git clone http://localhost/.../too-many-refs/.git
Cloning into 'too-many-refs'...
fatal: cannot exec 'fetch-pack': Argument list too long
50k tags is obviously an absurd number, but it is required to
demonstrate the problem on Linux because it has a much more generous
command line limit. On Windows the clone fails with as little as 500
tags in the above loop, which is getting uncomfortably close to the
number of tags you might see in real long lived repos.
This is not just theoretical, msysGit is already failing to clone our
company repo due to this. It's a large repo converted from CVS, nearly
10 years of history.
Four possible solutions were discussed on the Git mailing list (in no
particular order):
1) Call fetch-pack multiple times with smaller batches of refs.
This was dismissed as inefficient and inelegant.
2) Add option --refs-fd=$n to pass a an fd from where to read the refs.
This was rejected because inheriting descriptors other than
stdin/stdout/stderr through exec() is apparently problematic on Windows,
plus it would require changes to the run-command API to open extra
pipes.
3) Add option --refs-from=$tmpfile to pass the refs using a temp file.
This was not favored because of the temp file requirement.
4) Add option --stdin to pass the refs on stdin, one per line.
In the end this option was chosen as the most efficient and most
desirable from scripting perspective.
There was however a small complication when using stdin to pass refs to
fetch-pack. The --stateless-rpc option to fetch-pack also uses stdin for
communication with the remote server.
If we are going to sneak refs on stdin line by line, it would have to be
done very carefully in the presence of --stateless-rpc, because when
reading refs line by line we might read ahead too much data into our
buffer and eat some of the remote protocol data which is also coming on
stdin.
One way to solve this would be to refactor get_remote_heads() in
fetch-pack.c to accept a residual buffer from our stdin line parsing
above, but this function is used in several places so other callers
would be burdened by this residual buffer interface even when most of
them don't need it.
In the end we settled on the following solution:
If --stdin is specified without --stateless-rpc, fetch-pack would read
the refs from stdin one per line, in a script friendly format.
However if --stdin is specified together with --stateless-rpc,
fetch-pack would read the refs from stdin in packetized format
(pkt-line) with a flush packet terminating the list of refs. This way we
can read the exact number of bytes that we need from stdin, and then
get_remote_heads() can continue reading from the same fd without losing
a single byte of remote protocol data.
This way the --stdin option only loses generality and scriptability when
used together with --stateless-rpc, which is not easily scriptable
anyway because it also uses pkt-line when talking to the remote server.
Signed-off-by: Ivan Todoroski <grnch@gmx.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: fix tabbed preferences construction when using tcl 8.4
In 8.5 the incr command creates the target variable if it does not exist
but in 8.4 using incr on a non-existing variable raises an error. Ensure
we have created our counter variable when creating the tabbed dialog for
non-themed preferences.
Reported-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Portuguese Portuguese translations from Marco Sousa via Jiang Xin
* 'master' of git://github.com/git-l10n/git-po:
l10n: Add the Dutch translation team and initialize nl.po
l10n: Inital Portuguese Portugal language (pt_PT)
l10n: Improve zh_CN translation for Git 1.7.10-rc3
Make it clear that, when using commit --template, the message *must* be
changed or the commit will be aborted.
Helped-by: Junio C Hamano <gitster@pobox.com> Helped-by: Ivan Heffner <iheffner@gmail.com> Signed-off-by: Adam Monsen <haircut@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
commit: rephrase the error when user did not touch templated log message
When the user exited editor without editing the commit log template given
by "git commit -t <template>", the commit was aborted (correct) with an
error message that said "due to empty commit message" (incorrect).
This was because the original template support was done by piggybacking on
the check to detect an empty log message. Split the codepaths into two
independent checks to clarify the error.
commit: do not trigger bogus "has templated message edited" check
When "-t template" and "-F msg" options are both given (or worse yet,
there is "commit.template" configuration but a message is given in some
other way), the documentation says that template is ignored. However,
the "has the user edited the message?" check still used the contents of
the template file as the basis of the emptyness check.
These tests try to run "git commit" with various "forbidden" combinations
of options and expect the command to fail, but they do so without having
any change added to the index. We wouldn't be able to catch breakages
that would allow these combinations by mistake with them because the
command will fail with "nothing to commit" anyway.
Make sure we have something added to the index before running the command.
v1.7.9-8-g270a344 (config: stop using config_exclusive_filename) replaced
config_exclusive_filename with given_config_file. In one case this
resulted in a self-assignment, which is reported by clang as a warning.
Remove the useless code.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Why not just git grep -E ...?
That wouldn't work then the doubled words are separated by a newline.
This is derived from a Makefile syntax-check rule in gnulib's maint.mk:
http://git.sv.gnu.org/cgit/gnulib.git/tree/top/maint.mk
Signed-off-by: Jim Meyering <meyering@redhat.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* 'master' of git://github.com/git-l10n/git-po:
Add url of Swedish l10n team in TEAMS file
l10n: Review zh_CN translation for Git 1.7.10-rc1
Update Swedish translation (724t0f0u).
l10n: Update zh_CN translation for Git 1.7.10-rc1
l10n: Update git.pot (1 new message)
$COLUMNS must be unset to not interfere with the tests. The tests
already ignore the terminal size because output is redirected to a
file, but COLUMNS overrides terminal size detection and changes the
test output away from the standard 80.
Reported-by: Jeff King <peff@peff.net> Signed-off-by: Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
- Update for 1.7.10-rc1.
- Add a missing -e when generaring the "Untracked files" message.
- Fixed some wordings after playing with the localized version.
Documentation: improve description of GIT_EDITOR and preference order
Previously GIT_EDITOR was not listed in git(1) "Environment Variables" section,
which could be very confusing to users. Include it in "other" subsection along
with a link to git-var(1), since that is the page that fully describes all
places where editor can be set and also their preference order.
Also, git-var(1) did not say that hardcoded fallback 'vi' may have been changed
at build time. A user could be puzzled if 'nano' pops up even when none of the
mentioned environment vars or config.editor are set. Clarify this.
Ideally, the build system should be changed to reflect the chosen fallback
editor when creating the man pages. Not sure if that is even possible though.
Signed-off-by: Rodrigo Silva (MestreLion) <linux@rodrigosilva.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
documentation: fix alphabetic ordered list for git-rebase man page
An alphabetic ordered list (a.) is converted to numerical in
the man page (1.) so context comments naming 'a' were confusing,
fix that by not using ordered list notation for 'a' anb 'b' items.
Signed-off-by: Nelson Benitez Leon <nelsonjesus.benitez@seap.minhap.es> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If both la and context are zero at the start of the loop, la wraps around
and we end up reading from memory far away. Skip the loop in that case
instead.
Reported-by: Julia Lawall <julia.lawall@lip6.fr> Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge gitk changes from Paul Mackerras at git://ozlabs.org/~paulus/gitk
* git://ozlabs.org/~paulus/gitk:
gitk: Teach gitk to respect log.showroot
gitk: Add menu items for comparing a commit with the marked commit
gitk: Speed up resolution of short SHA1 ids
gitk: Use symbolic font names "sans" and "monospace" when available
gitk: Skip over AUTHOR/COMMIT_DATE when searching all fields
gitk: Make "git describe" output clickable, too
gitk: Fix the display of files when filtered by path
gitk: Use a tabbed dialog to edit preferences
gitk: Use "gitk: repo-top-level-dir" as window title
In early days, all projects managed by git (except for git itself) had the
product of a fairly mature development history in their first commit, and
it was deemed unnecessary clutter to show additions of these thousands of
paths as a patch.
"git log" learned to show the patch for the initial commit without requiring
--root command line option at 0f03ca9 (config option log.showroot to show
the diff of root commits, 2006-11-23).
Teach gitk to respect log.showroot.
[paulus@samba.org: Cleaned up the Tcl a bit, use --bool on the
git config call]
Signed-off-by: Marcus Karlsson <mk@acc.umu.se> Signed-off-by: Paul Mackerras <paulus@samba.org>
merge-recursive: don't detect renames of empty files
Merge-recursive detects renames so that if one side modifies
"foo" and the other side moves it to "bar", the modification
is applied to "bar". However, our rename detection is based
on content analysis, it can be wrong (i.e., two files were
not intended as a rename, but just happen to have the same
or similar content).
This is quite rare if the files actually contain content,
since two unrelated files are unlikely to have exactly the
same content. However, empty files present a problem, in
that there is nothing to analyze. An uninteresting
placeholder file with zero bytes may or may not be related
to a placeholder file with another name.
The result is that adding content to an empty file may cause
confusion if the other side of a merge removed it; your
content may end up in another random placeholder file that
was added.
Let's err on the side of caution and not consider empty
files as renames. This will cause a modify/delete conflict
on the merge, which will let the user sort it out
themselves.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
teach diffcore-rename to optionally ignore empty content
Our rename detection is a heuristic, matching pairs of
removed and added files with similar or identical content.
It's unlikely to be wrong when there is actual content to
compare, and we already take care not to do inexact rename
detection when there is not enough content to produce good
results.
However, we always do exact rename detection, even when the
blob is tiny or empty. It's easy to get false positives with
an empty blob, simply because it is an obvious content to
use as a boilerplate (e.g., when telling git that an empty
directory is worth tracking via an empty .gitignore).
This patch lets callers specify whether or not they are
interested in using empty files as rename sources and
destinations. The default is "yes", keeping the original
behavior. It works by detecting the empty-blob sha1 for
rename sources and destinations.
One more flexible alternative would be to allow the caller
to specify a minimum size for a blob to be "interesting" for
rename detection. But that would catch small boilerplate
files, not large ones (e.g., if you had the GPL COPYING file
in many directories).
A better alternative would be to allow a "-rename"
gitattribute to allow boilerplate files to be marked as
such. I'll leave the complexity of that solution until such
time as somebody actually wants it. The complaints we've
seen so far revolve around empty files, so let's start with
the simple thing.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The read-cache implementation defines this static function,
but it is a generally useful concept in git. Let's give
the empty blob the same treatment as the empty tree,
providing both hex and binary forms of the sha1.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This macro already evaluates to the correct type, as it
casts the string literal to "unsigned char *" itself
(and callers who want the literal can use the _LITERAL
form).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Add menu items for comparing a commit with the marked commit
Sometimes one wants to see the different between two commits that are
a long distance apart in the graph display. This is difficult to do
with the "Diff this -> selected" and "Diff selected -> this" menu
items because the need to maintain the selection means that one can't
use the find facilities or the reference list window to navigate from
one to the other.
This provides an alternative using the mark. Having found one commit,
one marks it with the "Mark this commit" menu item, then navigates to
the other commit and uses the new "Diff this -> marked commit" and/or
"Diff marked commit -> this" menu items.
Even though 1.7.9.x series does not open the editor by default
when merging in general, it does do so in one occassion: when
merging an annotated tag. And worse yet, there is no good way
for older scripts to decline this.
Backport the support for GIT_MERGE_AUTOEDIT environment variable
from 1.7.10 track to help those stuck on 1.7.9.x maintenance
track.
t4202: add test for "log --graph --stat -p" separator lines
Add tests to make sure that the three-dash separator lines appear
after the graph ancestry lines, and also the graph ancestry lines
are not broken between the diffstat and the patch.
Signed-off-by: Lucian Poston <lucian.poston@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
log --graph --stat: three-dash separator should come after graph lines
Output from "git log --graph --stat -p" emits the three-dash separator
line before the graph that shows ancestry lines. The separator should
come after the ancestry lines just like all the other output.
Signed-off-by: Lucian Poston <lucian.poston@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
On large repositories such as the Linux kernel, it can take quite a
noticeable time (several seconds) for gitk to resolve short SHA1 IDs
to their long form. This speeds up the process by maintaining lists
of IDs indexed by the first 4 characters of the SHA1 ID, speeding up
the search by a factor of 65536 on large repositories.
gitk: Use symbolic font names "sans" and "monospace" when available
The following only concerns systems using X and the client-side font
rendering framework from freedesktop.org. Windows and Mac OS X are
not affected.
Starting with version 8.5, Tk uses freetype and fontconfig by default
to render fonts on platforms that support it. Gitk currently defaults
to the font Helvetica for the interface and Courier for diffs, and
both unfortunately look rather bad on screen in the default
configuration on many Linux distros with anti-aliasing and poor
hinting.
It is better to default to "sans" and "monospace", which are mapped by
fontconfig to some appropriate font of the sysadmin and user's
choosing (typically Bitstream Vera Sans and Mono). The result looks
more sensible and it makes gitk feel like a well-behaved software
citizen since its fonts match other native apps.
This patch does not change the appearance of gitk for users that have
already run it, since gitk uses the remembered UI and diff font names
from ~/.gitk.
Requested-by: Michael Biebl <biebl@debian.org> Reviewed-by: Josh Triplett <josh@joshtriplett.org> Acked-by: Mark Hills <mark@pogo.org.uk> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Skip over AUTHOR/COMMIT_DATE when searching all fields
This prevents a search for a number like "105" on "All Fields" from
matching against the raw author and commit timestamps. These
timestamps were already not searchable by themselves, and the
displayed format does not match the query string anyway.
Signed-off-by: Frédéric Brière <fbriere@fbriere.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Automake's contribution guidelines suggest using "git describe" output
in commit logs to reference previous commits. By contrast, in
coreutils, I had acquired the habit of using a bare SHA1 prefix (8 hex
digits), since gitk creates clickable links for that, and not for "git
describe" output.
I prefer the readability of the full "git describe" output, yet want
to retain the gitk links, so this renders as clickable not just
SHA1-like strings, but also an SHA1-like string that is prefixed by
"-g".
Signed-off-by: Jim Meyering <meyering@redhat.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Fix the display of files when filtered by path
Launching 'gitk -- .' or 'gitk -- ..\t' restricts the display to files
under the given directory but the file list is left empty. This is because
the path_filter function fails to match the filenames which are relative
to the working tree to the filter which is filessytem relative.
This solves the problem by making both names fully qualified filesystem
paths before performing the comparison.
Tested-by: David Aguilar <davvid@gmail.com> Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net> Signed-off-by: Paul Mackerras <paulus@samba.org>
This commit converts the user preferences dialog into a tabbed property
sheet grouping general properties, colours and font selections onto
separate pages. The previous implementation was exceeding the screen
height on some systems and this avoids such problems and permits extension
using new pages in the future.
If themed Tk is unavailable or undesired a reasonable facsimile of the
tabbed notebook widget is used instead.
Signed-off-by: Pat Thoyts <patthoyts@users.sourceforge.net> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Use "gitk: repo-top-level-dir" as window title
Previously, when run in a subdirectory, gitk would show the name
of this subdirectory as title, which was misleading. When run with
GIT_DIR set, it would show the cwd, which is even more misleading.
In case of non-bare repos, the .git suffix in the path is skipped.
Signed-off-by: Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl>
perl/Makefile: install Git::I18N under NO_PERL_MAKEMAKER
When I added the i18n infrastructure in v1.7.8-rc2-1-g5e9637c I forgot
to install Git::I18N also when NO_PERL_MAKEMAKER=YesPlease was
set. Change the generation of the fallback perl.mak file to do that.
Now Git/I18N.pm is installed alongside Git.pm in such a way that
anything that uses GITPERLLIB will find it.
Reported-by: Tom G. Christensen <tgc@statsbiblioteket.dk> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code to validate the history connectivity between old refs and new
refs used by fetch and receive-pack, introduced in 1.7.8, was grossly
inefficient and unnecessarily tried to re-validate integrity of individual
objects. This essentially reverts that performance regression.
Merge "two fixes for fast-import's 'ls' command" from Jonathan
Andrew Sayers noticed that the svn-fe | git fast-import pipeline
mishandles a subversion history that copies the root directory to a
sub-directory (e.g. doing `svn cp . trunk` to standardise your
layout). As David Barr explained, the bug arises when the following
command is sent to git fast-import:
'ls' SP ':1' SP LF
Instead of reading back what is at the root of r1, it unconditionally
reports the path as missing.
After sleeping on it, here are two patches for 'maint'. One plugs a
memory leak. The other ensures that trying to pass an empty path to
the 'ls' command results in an error message that can help the
frontend author instead of the silently broken conversion Andrew
found.
Then we can carefully add 'ls ""' support in 1.7.11.
* commit 'refs/pull-request-tags/jn/maint-fast-import-empty-ls':
fast-import: don't allow 'ls' of path with empty components
fast-import: leakfix for 'ls' of dirty trees
* th/git-diffall:
contrib/diffall: fix cleanup trap on Windows
contrib/diffall: eliminate duplicate while loops
contrib/diffall: eliminate use of tar
contrib/diffall: create tmp dirs without mktemp
contrib/diffall: comment actual reason for 'cdup'
Git 1.7.8 introduced an object and history re-validation step after
"fetch" or "push" causes new history to be added to a receiving
repository. This is to protect a malicious server or pushing client from
corrupting the repository by taking advantage of an existing corrupt
object that is unconnected to existing history.
But this check is way over-pessimistic. During "fetch" or "receive-pack"
(the server side of "push"), unpack-objects and index-pack already
validate individual objects that are received, and the only thing we would
want to catch are corrupted objects that already happen to exist in our
repository but are not referenced from our refs. Such objects must have
been written by an earlier run of our codepaths that write out loose
objects or packfiles, and they must have done the validation of individual
objects when they did so. The only thing left to worry about is the
connectivity integrity, which can be checked with "rev-list --objects",
which is much cheaper. We have been paying the 5x to 8x runtime overhead
the --verify-objects often adds for no real gain.
Revert check_everything_connected() not to use this over-pessimistic
check.
Credit goes to Nguyễn Thái Ngọc Duy, who originally identified the
performance regression and endured multiple rounds of reviews to fix it.
clean: preserve nested git worktree in subdirectories
remove_dir_recursively() has a check to avoid removing the directory it
was asked to remove without recursing into it and report success when the
directory is the top level of a working tree of a nested git repository,
to protect such a repository from "clean -f" (without double -f). If a
working tree of a nested git repository is in a subdirectory of a toplevel
project, however, this protection did not apply by mistake; we forgot to
pass the REMOVE_DIR_KEEP_NESTED_GIT down to the recursive removal
codepath.
This requires us to also teach the higher level not to remove the
directory it is asked to remove, when the recursed invocation did not
remove the directory it was asked to remove due to a nested git
repository, as it is not an error to leave the parent directories of such
a nested repository.
notes-merge: Don't remove .git/NOTES_MERGE_WORKTREE; it may be the user's cwd
When a manual notes merge is committed or aborted, we need to remove the
temporary worktree at .git/NOTES_MERGE_WORKTREE. However, removing the
entire directory is not good if the user ran the 'git notes merge
--commit/--abort' from within that directory. On Windows, the directory
removal would simply fail, while on POSIX systems, users would suddenly
find themselves in an invalid current directory.
Therefore, instead of deleting the entire directory, we delete everything
_within_ the directory, and leave the (empty) directory in place.
This would cause a subsequent notes merge to abort, complaining about a
previous - unfinished - notes merge (due to the presence of
.git/NOTES_MERGE_WORKTREE), so we also need to adjust this check to only
trigger when .git/NOTES_MERGE_WORKTREE is non-empty.
Finally, adjust the t3310 manual notes merge testcases to correctly handle
the existence of an empty .git/NOTES_MERGE_WORKTREE directory.
Inspired-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
notes-merge: use opendir/readdir instead of using read_directory()
notes_merge_commit() only needs to list all entries (non-recursively)
under a directory, which can be easily accomplished with
opendir/readdir and would be more lightweight than read_directory().
read_directory() is designed to list paths inside a working
directory. Using it outside of its scope may lead to undesired effects.
Apparently, one of the undesired effects of read_directory() is that it
doesn't deal with being given absolute paths. This creates problems for
notes_merge_commit() when git_path() returns an absolute path, which
happens when the current working directory is in a subdirectory of the
.git directory.
Originally-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Updated-by: Johan Herland <johan@herland.net> Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t3310: illustrate failure to "notes merge --commit" inside $GIT_DIR/
The 'git notes merge' command expected to be run from the working
tree of the project being annotated, and did not anticipate getting
run inside $GIT_DIR/.
However, because we use $GIT_DIR/NOTES_MERGE_WORKTREE as a temporary
working space for the user to work on resolving conflicts, it is not
unreasonable for a user to run "git notes merge --commit" there. But
the command fails to do so.
Found-by: David Bremner <david@tethera.net> Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
remove_dir_recursively(): Add flag for skipping removal of toplevel dir
Add the REMOVE_DIR_KEEP_TOPLEVEL flag to remove_dir_recursively() for
deleting everything inside the given directory, but _not_ the given
directory itself.
Note that this does not pass the REMOVE_DIR_KEEP_NESTED_GIT flag, if set,
to the recursive invocations of remove_dir_recursively(). It is likely to
be a a bug that has been present since REMOVE_DIR_KEEP_NESTED_GIT was
introduced (a0f4afb), but this commit keeps the same behaviour for now.
Signed-off-by: Johan Herland <johan@herland.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prior to this commit, the cleanup trap that removes the tmp dir
created by the script would fail on Windows. The error was silently
ignored by the script.
On Windows, a directory cannot be removed while it is the working
directory of the process (thanks to Johannes Sixt on the Git list
for this info [1]).
This commit eliminates the 'cd' into the tmp directory that caused
the error.
The 'tar' utility is not available on all platforms (some only support
'gnutar'). An earlier commit created a work-around for this problem,
but a better solution is to eliminate the use of 'tar' completely.
Signed-off-by: Tim Henigan <tim.henigan@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff: tweak a _copy_ of diff_options with word-diff
When using word diff, the code sets the word_regex from various
defaults if it was not set already. The problem is that it does this
on the original diff_options, which will also be used in subsequent
diffs.
This means that when the word_regex is not given on the command line,
only the first diff for which a setting for word_regex (either from
attributes or diff.wordRegex) ever takes effect. This value then
propagates to the rest of the diff runs and in particular prevents
further attribute lookups.
Fix the problem of changing diff state once and for all, by working
with a _copy_ of the diff_options.
Noticed-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff: refactor the word-diff setup from builtin_diff_cmd
Quite a chunk of builtin_diff_cmd deals with word-diff setup, defaults
and such. This makes the function a bit hard to read, but is also
asymmetric because the corresponding teardown lives in free_diff_words_data
already.
Refactor into a new function init_diff_words_data. For simplicity,
also shuffle around some functions it depends on.
Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t4034: diff.*.wordregex should not be "sticky" in --word-diff
The test case applies a custom wordRegex to one file in a diff, and expects
that the default word splitting applies to the second file in the diff.
But the custom wordRegex is also incorrectly used for the second file.
Helped-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>