This was added long ago as part of the userdiff refactoring
for textconv, as internally it made the code simpler and
cleaner. However, there was never a concrete use case for
actually using the config variable.
Now that Matthieu Moy has provided such a use case, it's
easy to explain it using his example.
Signed-off-by: Jeff King <peff@peff.net> Acked-by: Matthieu Moy <Matthieu.Moy@grenoble-inp.fr> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If svn is built against one version of SQLite and run against another,
libsvn_subr needlessly errors out in operations that need to make a
commit.
That is clearly not a bug in git but let us consider the ramifications for
the test suite. git-svn uses libsvn directly and is probably broken by
that bug; it is right for git-svn tests to fail. The vcs-svn lib, on the
other hand, does not use libsvn and the test t9010 only uses svn to check
its work. This points to two possible improvements:
- do not disable most vcs-svn tests if svn is missing.
- skip validation rather than failing it when svn fails.
Bring about both by putting the svn invocations into a single test that
builds a repo to compare the test-svn-fe result against. The test will
always pass but only will set the new SVNREPO test prereq if svn succeeds;
and validation using that repo gets an SVNREPO prerequisite so it only
runs with working svn installations.
Works-around: http://bugs.debian.org/608925 Noticed-by: A Large Angry SCM <gitzilla@gmail.com> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
submodule: fix relative url parsing for scp-style origin
The function resolve_relative_url was not prepared to deal with an
scp-style origin 'user@host:path' in the case where 'path' is only a
single component. Fix this by extending the logic that strips one
path component from the $remoteurl.
Also add tests for both styles of URLs.
Noticed-by: Jeffrey Phillips Freeman <jeffrey.freeman@syncleus.com> Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/githooks: post-rewrite-copy-notes never existed
The documentation for the post-rewrite hook contains a paragraph from
its early development, where the automatic notes copying facilities
were not part of the series and thus this had to be a hook. Later
versions of the series implemented notes copying as a core feature.
Thus mentioning post-rewrite-copy-notes was never correct. As the
other hooks do not have a "there is no default hook, but..." sentence
unless they ship a sample hook in either templates or contrib, we
simply remove the whole paragraph.
Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The --worktree-attributes option was correctly documented in ba053ea
(archive: do not read .gitattributes in working directory,
2009-04-18). However, later in 9b4c8b0 (archive documentation:
attributes are taken from the tree by default, 2010-02-10) the
misspelling "--work-tree-attributes" was used to refer to it. Fix
this.
Noticed-by: Jeffrey Phillips Freeman <jeffrey.freeman@syncleus.com> Signed-off-by: Thomas Rast <trast@student.ethz.ch> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The BSLASHPSPEC tests (11-13) fail on cygwin, since you can't
create files containing an backslash character in the name.
In order to skip these tests, we simply stop (incorrectly)
asserting the BSLASHPSPEC prerequisite in test-lib.sh.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t3032-*.sh: Do not strip CR from line-endings while grepping on MinGW
By default grep reads in text mode and converts CRLF into LF line
endings, which causes tests 4, 6 and 8 to fail. In a similar manner
to commit a94114ad (Do not strip CR when grepping HTTP headers,
2010-09-12), we set (and export) the GREP_OPTIONS variable to -U so
that grep will use binary mode.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t3032-*.sh: Pass the -b (--binary) option to sed on cygwin
The test using the conflict_hunks helper function (test 9) fails
on cygwin, since sed (by default) throws away the CR from CRLF
line endings. This behaviour is undesirable, since the validation
code expects the CRLF line-ending to be present. In order to fix
the problem we pass the -b (--binary) option to sed, using the
SED_OPTIONS variable. We use the SED_STRIPS_CR prerequisite in the
conditional initialisation of SED_OPTIONS.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t6038-*.sh: Pass the -b (--binary) option to sed on cygwin
The tests using the fuzz_conflict helper function (tests 5-6)
fail on cygwin in the same way they used to on MinGW, prior
to commit ca02ad3. The solution is also the same; passing the
-b (--binary) option to sed, using the SED_OPTIONS variable.
We introduce a new prerequisite SED_STRIPS_CR to use in the
conditional initialisation of SED_OPTIONS, rather than MINGW.
The new prerequisite is set in test-lib.sh for both MinGW and
Cygwin.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Acked-by: Johannes Sixt <j6t@kdbg.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The executable bit on gitk-git/gitk was lost (accidentally it seems) by
commit 62ba5143ec2ab9d4083669b1b1679355e7639cd5. Put it back, so that
gitk can be run directly from a git.git checkout.
Note that the script is already executable in gitk.git, just not in
git.git.
Signed-off-by: Anders Kaseorg <andersk@mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
'git-remote add' creates a remote.origin.fetch entry in the config, we
want to replace this entry rather than add another one (which will
cause 'git fetch' to error).
This adds 'git config --remove-section remote.origin' after the fetch
for encouraging users to only use "git svn" for future updates.
[ew: rewording of commit message for present tense]
Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: StephenB <mail4stb@gmail.com>
On Windows, system() executes with cmd.exe instead of /bin/sh. This
means that aliases currently has to be batch-scripts instead of
bourne-scripts. On top of that, cmd.exe does not handle single quotes,
which is what the code-path currently uses to handle arguments with
spaces.
To solve both problems in one go, use run_command_v_opt() to execute
the alias. It already does the right thing prepend "sh -c " to the
alias.
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com> Reviewed-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
POSIX leaves as unspecified the handling of labels greater than 8
characters. Apparently, Sun decided to treat them as errors. Make sed on
Solaris happy by trimming the length of labels to 8 characters.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t0001,t1510,t3301: use sane_unset which always returns with status 0
On some shells (like /usr/xpg4/bin/sh on Solaris), unset will exit
non-zero when passed the name of a variable that has not been set. Use
sane_unset instead so that the return value of unset can be ignored while
the && linkage of the test script can be preserved.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Junio C Hamano <gitster@pobox.com>
GNU printf, and many others, will print the string "(null)" if a NULL
pointer is passed as the argument to a "%s" format specifier. Some
implementations (like on Solaris) do not detect a NULL pointer and will
produce a segfault in this case.
So, fix this by ensuring that pointer variables do not contain the value
NULL. Assign the string "(null)" to the variables are NULL.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitweb: remove unnecessary test when closing file descriptor
It happens that closing file descriptor fails whereas the blob is
perfectly readable. According to perlman the reasons could be:
If the file handle came from a piped open, "close" will additionally
return false if one of the other system calls involved fails, or if the
program exits with non-zero status. (If the only problem was that the
program exited non-zero, $! will be set to 0.) Closing a pipe also waits
for the process executing on the pipe to complete, in case you want to
look at the output of the pipe afterwards, and implicitly puts the exit
status value of that command into $?.
Prematurely closing the read end of a pipe (i.e. before the process writ-
ing to it at the other end has closed it) will result in a SIGPIPE being
delivered to the writer. If the other end can't handle that, be sure to
read all the data before closing the pipe.
In this case we don't mind that close fails.
Signed-off-by: Sylvain Rabot <sylvain@abstraction.fr> Signed-off-by: Junio C Hamano <gitster@pobox.com>
lib-git-svn.sh: Move web-server handling code into separate function
This library file is currently sourced by 57 test files, of which
only four may (optionally) start a web-server in order to access
the svn repo via an http url, rather than a file url.
In addition to isolating the current web-server handling code from
the majority of tests, in a new prepare_httpd function, we also
add some more error checking and reporting code to validate the
apache installation. Only those tests which attempt to start the
web-server, by calling start_httpd, will execute this code.
Note that it is important for start_httpd to return an error
indication, if prepare_httpd fails, so that the failure to use
the web-server, as requested by the user, should not go unnoticed.
(Unless the svnrepo variable is set to an http url at the end of
start_httpd, the remaining tests will use file urls, without
comment.)
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cvsimport: handle the parsing of uppercase config options
The current code leads to
fatal: bad config value for 'cvsimport.r' in .git/config
for a standard use case with cvsimport.r set.
cvsimport sets internal variables by checking the config for each
possible command line option. The problem is that config items are case
insensitive, so config.r and config.R are the same. The ugly error is
due to that fact that cvsimport expects a bool for -R (and thus
config.R) but a remote name for -r (and thus config.r).
Fix this by making cvsimport expect long names for uppercase options.
config options for cvsimport have been undocumented so far, though
present in the code and advertised in several tutorials. So one may read
"enhance" for "fix". Similarly, the names for the options are
"documented" in the code, waitiing for their lowercase equivalents to be
transformed into long config options, as well.
Helped-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
unpack_trees(): skip trees that are the same in all input
unpack_trees() merges two trees (the current HEAD and the destination
commit) when switching to another branch, checking and updating the index
entry where the destination differs from the current HEAD. It merges three
trees (the common ancestor, the current HEAD and the other commit) when
performing a three-way merge, checking and updating the index entry when
the merge result differs from the current HEAD. It does so by walking the
input trees in parallel all the way down to the leaves.
One common special case is a directory is identical across the trees
involved in the merge. In such a case, we do not have to descend into the
directory at all---we know that the end result is to keep the entries in
the current index.
This optimization cannot be applied in a few special cases in
unpack_trees(), though. We need to descend into the directory and update
the index entries from the target tree in the following cases:
- When resetting (e.g. "git reset --hard"); and
- When checking out a tree for the first time into an empty working tree
(e.g. "git read-tree -m -u HEAD HEAD" with missing .git/index).
Ever since v1.7.4-rc0~125^2~8 (daemon: use run-command api for async
serving, 2010-11-04), git daemon spawns child processes instead of
forking to serve requests. The child processes learn that they are
being run for this purpose from the presence of the --serve command
line flag.
When running with <ok_path> arguments, the --serve flag is treated
as one of the path arguments and the special child behavior does
not kick in. So the child becomes an ordinary git daemon process,
notices that all the addresses it needs are in use, and exits with
the message "fatal: unable to allocate any listen sockets on port
9418".
Fix it by putting --serve at the beginning of the command line,
where the flag cannot be mistaken for a path argument.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Acked-by: Erik Faye-Lund <kusmabite@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
While at it, use $cgi->img to simplify this code. (CGI.pm learned
most HTML4 tags by version 2.79, so this should be portable to perl
5.8, though I haven't tested.)
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
With v1.5.0-rc0~169 (gitweb: Fix Atom feed <logo>: it is $logo,
not $logo_url, 2006-12-04), the logo URI to be written to Atom
feeds was corrected but the case of no logo forgotten.
Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Acked-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Merge branch 'ks/blame-worktree-textconv-cached' into maint
* ks/blame-worktree-textconv-cached:
fill_textconv(): Don't get/put cache if sha1 is not valid
t/t8006: Demonstrate blame is broken when cachetextconv is on
t9001: use older Getopt::Long boolean prefix '--no' rather than '--no-'
The '--no-chain-reply-to' option is a Getopt::Long boolean option. The
'--no-' prefix (as in --no-chain-reply-to) for boolean options is not
supported in Getopt::Long version 2.32 which was released with Perl 5.8.0.
This version only supports '--no' as in '--nochain-reply-to'. More recent
versions of Getopt::Long, such as version 2.34, support either prefix. So
use the older form in the tests.
* nd/setup: (47 commits)
setup_work_tree: adjust relative $GIT_WORK_TREE after moving cwd
git.txt: correct where --work-tree path is relative to
Revert "Documentation: always respect core.worktree if set"
t0001: test git init when run via an alias
Remove all logic from get_git_work_tree()
setup: rework setup_explicit_git_dir()
setup: clean up setup_discovered_git_dir()
t1020-subdirectory: test alias expansion in a subdirectory
setup: clean up setup_bare_git_dir()
setup: limit get_git_work_tree()'s to explicit setup case only
Use git_config_early() instead of git_config() during repo setup
Add git_config_early()
git-rev-parse.txt: clarify --git-dir
t1510: setup case #31
t1510: setup case #30
t1510: setup case #29
t1510: setup case #28
t1510: setup case #27
t1510: setup case #26
t1510: setup case #25
...
Fix false positives in t3404 due to SHELL=/bin/false
If the user's shell in NSS passwd is /bin/false (eg as found during Gentoo's
package building), the git-rebase exec tests will fail, because they call
$SHELL around the command, and in the existing testcase, $SHELL was not being
cleared sufficently.
This lead to false positive failures of t3404 on systems where the package
build user was locked down as noted above.
Signed-off-by: "Robin H. Johnson" <robbat2@gentoo.org>
X-Gentoo-Bug: 349083
X-Gentoo-Bug-URL: http://bugs.gentoo.org/show_bug.cgi?id=349083 Signed-off-by: Junio C Hamano <gitster@pobox.com>
The default function name discovery already works quite well for Perl
code... with the exception of here-documents (or rather their ending).
sub foo {
print <<END
here-document
END
return 1;
}
The default funcname pattern treats the unindented END line as a
function declaration and puts it in the @@ line of diff and "grep
--show-function" output.
With a little knowledge of perl syntax, we can do better. You can
try it out by adding "*.perl diff=perl" to the gitattributes file.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
setup_work_tree: adjust relative $GIT_WORK_TREE after moving cwd
When setup_work_tree() is called, it moves cwd to $GIT_WORK_TREE and
makes internal copy of $GIT_WORK_TREE absolute. The environt variable,
if set by user, remains unchanged. If the variable is relative, it is
no longer correct because its base dir has changed.
Instead of making $GIT_WORK_TREE absolute too, we just say "." and let
subsequent git processes handle it.
Reported-by: Michel Briand <michelbriand@free.fr> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
rebase --skip: correctly wrap-up when skipping the last patch
When "rebase --skip" is used to skip the last patch in the series, the
code to wrap up the rewrite by copying the notes from old to new commits
and also by running the post-rewrite hook was bypassed.
* rj/maint-test-fixes:
t9501-*.sh: Fix a test failure on Cygwin
lib-git-svn.sh: Add check for mis-configured web server variables
lib-git-svn.sh: Avoid setting web server variables unnecessarily
t9142: Move call to start_httpd into the setup test
t3600-rm.sh: Don't pass a non-existent prereq to test #15
* nd/maint-fix-add-typo-detection:
Revert "excluded_1(): support exclude files in index"
unpack-trees: fix sparse checkout's "unable to match directories"
unpack-trees: move all skip-worktree checks back to unpack_trees()
dir.c: add free_excludes()
cache.h: realign and use (1 << x) form for CE_* constants
Add some tests to document the correct behavior of (possibly aliased)
init when run within and outside a git directory.
If I set up a simple git alias “quietinit = init --quiet”, usually it
will work just like ‘git init --quiet’.
There are some differences, unfortunately, since in the process of
checking for aliases, git has to look for a .git/config file. If ‘git
quietinit’ is run from a subdirectory of an existing git repository,
that repository’s configuration will affect the configuration of the
new repository. In particular, the new repository can inherit
bogus values for core.bare and core.worktree.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This logic is now only used by cmd_init_db(). setup_* functions do not
rely on it any more. Move all the logic to cmd_init_db() and turn
get_git_work_tree() into a simple function.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function is the most complex one among the three setup_*
functions because all GIT_DIR, GIT_WORK_TREE, core.worktree and
core.bare are involved.
Because core.worktree is only effective inside
setup_explicit_git_dir() and the extra code in setup_git_directory()
is to handle that. The extra code can now be retired.
Also note that setup_explicit assignment is removed, worktree setting
is no longer decided by get_git_work_tree(). get_git_work_tree() will
be simplified in the next commit.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
setup: limit get_git_work_tree()'s to explicit setup case only
get_git_work_tree() takes input as core.worktree, core.bare,
GIT_WORK_TREE and decides correct worktree setting.
Unfortunately it does not do its job well. core.worktree and
GIT_WORK_TREE should only be taken into account, if GIT_DIR is set
(which is handled by setup_explicit_git_dir). For other setup cases,
only core.bare matters.
Add a temporary variable setup_explicit to adjust get_git_work_tree()
behavior as such. This variable will be gone once setup_* rework is
done.
Also remove is_bare_repository_cfg check in set_git_work_tree() to
ease the rework. We are going to check for core.bare and core.worktree
early before setting worktree. For example, if core.bare is true, no
need to set worktree.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This version of git_config() will be used during repository setup.
As a repository is being set up, $GIT_DIR is not nailed down yet,
git_pathdup() should not be used to get $GIT_DIR/config.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Filtering to support keyword expansion may need the name of
the file being filtered. In particular, to support p4 keywords
like
$File: //depot/product/dir/script.sh $
the smudge filter needs to know the name of the file it is
smudging.
Allow "%f" in the custom filter command line specified in the
configuration. This will be substituted by the filename
inside a single-quote pair to be passed to the shell.
Signed-off-by: Pete Wyckoff <pw@padd.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Improve error messages when temporary file creation fails
Before, when creating a temporary file failed, a generic 'Unable to create
temporary file' message was printed. In some cases this could lead to
confusion as to which directory should be checked for correct permissions etc.
This patch adds the template for the temporary filename to the error message,
converting it to an absolute path if needed. A test verifies that the template
is indeed printed when pointing to a nonexistent or unwritable directory.
A copy of the original template is made in case mkstemp clears the template.
Signed-off-by: Arnout Engelen <arnouten@bzzt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* ks/blame-worktree-textconv-cached:
fill_textconv(): Don't get/put cache if sha1 is not valid
t/t8006: Demonstrate blame is broken when cachetextconv is on
* nd/oneline-sha1-name-from-specific-ref:
get_sha1: handle special case $commit^{/}
get_sha1: support $commit^{/regex} syntax
get_sha1_oneline: make callers prepare the commit list to traverse
get_sha1_oneline: fix lifespan rule of temp_commit_buffer variable
Unlike bash and ksh, dash passes through hexadecimal \xcc escapes.
So when run with dash, these tests *pass* (since '\xcc' is a perfectly
reasonable filename) but they are not testing what was intended.
Use octal escapes instead, in the spirit of v1.6.1-rc1~55^2
(2008-11-09).
Reported-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some shells, for example dash versions older than 0.5.4, need to
spell a variable reference as '$N' rather than 'N' in an arithmetic
expansion. In order to avoid the syntax error, we change the
offending variable reference from 'i' to '$i' in function scramble.
There is nothing bash specific to this test script (and we shouldn't
have any bash dependent test). Fix its shebang line.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
am --abort: keep unrelated commits since the last failure and warn
After making commits (either by pulling or doing their own work) after a
failed "am", the user will be reminded by next "am" invocation that there
was a failed "am" that the user needs to decide to resolve or to get rid
of the old "am" attempt. The "am --abort" option was meant to help the
latter. However, it rewinded the HEAD back to the beginning of the failed
"am" attempt, discarding commits made (perhaps by mistake) since.
When determine_author_info() returns to the calling prepare_to_commit(),
we already know the pieces of information necessary to determine what
author ident will be used in the final message, but deferred making a call
to fmt_ident() before the final commit_tree(). Most importantly, we would
open the editor to ask the user to compose the log message before it.
As one important side effect of fmt_ident() is to error out when the given
information is malformed, this resulted in us spawning the editor first
and then refusing to commit due to error, even though we had enough
information to detect the error before starting the editor, which was
annoying.
Move the fmt_ident() call to the end of determine_author_info() where we
have final determination of author info to rectify this.
Quite a few configuration variables have been added since 226b343
(completion: add missing configuration variables to _git_config(),
2009-05-03). Add these variables to the Bash completion script.
Also remove the obsolete 'add.ignore-errors' and
'color.grep.external', as well as 'diff.renameLimit.', which never
existed and rename the misspelled 'sendemail.aliasesfiletype'.
Signed-off-by: Martin von Zweigbergk <martin.von.zweigbergk@gmail.com> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the user gives "git commit --date=foobar", we silently
ignore the --date flag. We should note the error.
This patch puts the fix at the lowest level of fmt_ident,
which means it also handles GIT_AUTHOR_DATE=foobar, as well.
There are two down-sides to this approach:
1. Technically this breaks somebody doing something like
"git commit --date=now", which happened to work because
bogus data is the same as "now". Though we do
explicitly handle the empty string, so anybody passing
an empty variable through the environment will still
work.
If the error is too much, perhaps it can be downgraded
to a warning?
2. The error checking happens _after_ the commit message
is written, which can be annoying to the user. We can
put explicit checks closer to the beginning of
git-commit, but that feels a little hack-ish; suddenly
git-commit has to care about how fmt_ident works. Maybe
we could simply call fmt_ident earlier?
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
fill_textconv(): Don't get/put cache if sha1 is not valid
When blaming files in the working tree, the filespec is marked with
!sha1_valid, as we have not given the contents an object name yet. The
function to cache textconv results (keyed on the object name), however,
didn't check this condition, and ended up on storing the cached result
under a random object name.
Cc: Axel Bonnet <axel.bonnet@ensimag.imag.fr> Cc: Clément Poulain <clement.poulain@ensimag.imag.fr> Cc: Diane Gasselin <diane.gasselin@ensimag.imag.fr> Cc: Jeff King <peff@peff.net> Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t/t8006: Demonstrate blame is broken when cachetextconv is on
I have a git repository with lots of .doc and .pdf files. There diff
works ok, but blaming is painfully slow without textconv cache, and with
textconv cache, blame says lots of lines are 'Not Yet Committed' which
is wrong.
Here is a test that demonstrates the problem.
Cc: Axel Bonnet <axel.bonnet@ensimag.imag.fr> Cc: Clément Poulain <clement.poulain@ensimag.imag.fr> Cc: Diane Gasselin <diane.gasselin@ensimag.imag.fr> Cc: Jeff King <peff@peff.net> Signed-off-by: Kirill Smelkov <kirr@landau.phys.spbu.ru> Signed-off-by: Junio C Hamano <gitster@pobox.com>