On Sat, 15 Mar 2008, SZEDER G?bor wrote:
>
> The testcase usually fails during the first 25 run, but sometimes it
> runs more than 100 times before failing.
Damn, this series has had more subtle issues than I ever expected.
'git stash' creates its saved working tree object with:
# state of the working tree
w_tree=$( (
rm -f "$TMP-index" &&
cp -p ${GIT_INDEX_FILE-"$GIT_DIR/index"} "$TMP-index" &&
GIT_INDEX_FILE="$TMP-index" &&
export GIT_INDEX_FILE &&
git read-tree -m $i_tree &&
git add -u &&
git write-tree &&
rm -f "$TMP-index"
) ) ||
die "Cannot save the current worktree state"
which creates a new index file with the updates, and writes the tree from
that.
We have this logic where we compare the timestamp of the index with the
timestamp of the files and we then write them out "smudged" if they are
the same, and it basically depends on the fact that the date on the index
file is compared with the date encoded in the stat information itself.
And what is going on is:
- we create a new index file with that "cp". We are careful to preserve
the timestamps by using "-p", so this one should be all ok.
- then we *update* that index by resetting it to the tree with git
read-tree, but now we do *not* preserve the timestamp on this new copy
any more, even though we copy over all the timestamps on the files that
are indexed from the stat information!
Now, we always had that problem when re-writing the index, but we had this
clever workaround in the writing part: if the source had racily clean
entries, then when we wrote those out (and thus can't depend on the index
fiel timestamp showing that they are racily clean any more!), we would
smudge them when writing.
IOW, we handle this issue by having write_index() do this:
for (i = 0; i < entries; i++) {
...
if (is_racy_timestamp(istate, ce))
ce_smudge_racily_clean_entry(ce);
..
when writing out entries. And that all took care of it, because now when
we wrote the new index, we'd change the timestamp on the index, yes, but
we'd smudge the entries we wrote out, so now the resulting index would
still show that file as not-up-to-date any more.
But with commit 34110cd4e3 ("Make
'unpack_trees()' have a separate source and destination index"), this
logic no longer triggers, because we now write out the "result" index, and
that one never got its timestamp updated from the source index, so it had
lost all that "is_racy_timestamp()" information!
This trivial patch fixes it. It looks trivial, and it's a simple fix, but
boy did it take me way too much thinking and explaining to myself to
explain why there was a problem in the first place!
The trivial fix is to just copy the index timestamp from the source index
into the result index. But we only do this if we *have* a source index, of
course, and if we will even bother to use the result.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://git.kernel.org/pub/scm/gitk/gitk:
gitk: initial Italian translation
gitk: Default to using po2msg.sh if msgfmt doesn't grok --tcl, -l and -d
gitk: Avoid Tcl error when switching views
[PATCH] gitk: Don't show local changes when we there is no work tree
[PATCH] gitk: Add horizontal scrollbar to the diff view
[PATCH] gitk: make autoselect optional
[PATCH] gitk: Mark another string for translation
[PATCH] Add an --argscmd flag to get the list of refs to show
gitk: Only restore window size from ~/.gitk, not position
This is a similar change to that submitted by Junio C Hamano for
git-gui. It tests whether the msgfmt command can be run successfully
with --tcl, -l and -d, and if not, falls back to using po/po2msg.sh.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Currently "git web--browse" is restricted to a set of commands defined
in the script. You can subvert the "browser.<tool>.path" to force "git
web--browse" to use a different command, but if you have a command
whose invocation syntax does not match one of the current tools then
you would have to write a wrapper script for it.
This patch adds a git config variable "browser.<tool>.cmd" which
allows a more flexible browser choice.
If you run "git web--browse" with -t/--tool, -b/--browser or the
"web.browser" config variable set to an unrecognized tool then "git
web--browse" will query the "browser.<tool>.cmd" config variable. If
this variable exists, then "git web--browse" will treat the specified
tool as a custom command and will use a shell eval to run the command
with the URLs added as extra parameters.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
merge-file: handle empty files gracefully
merge-recursive: handle file mode changes
Minor wording changes in the keyboard descriptions in git-add --interactive.
git fetch: Take '-n' to mean '--no-tags'
quiltimport: fix misquoting of parsed -p<num> parameter
git-quiltimport: better parser to grok "enhanced" series files.
read-tree -m can read up to MAX_TREES, which was arbitrarily set to 8 since
August 2007 (4 is needed to deal with 2 merge-base case).
However, the updated unpack_trees() code had an advertised limit of 4
(which it enforced). In reality the code was prepared to take only 3
trees and giving 4 caused it to stomp on its stack. Rename the MAX_TREES
constant to MAX_UNPACK_TREES, move it to the unpack-trees.h common header
file, and use it from both places to avoid future confusion.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Earlier, it would error out while trying to read and/or writing them.
Now, calling merge-file with empty files is neither interesting nor
useful, but it is a bug that needed fixing.
Noticed by Clemens Buchacher.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
File mode changes should be handled similarly to changes of content.
That is, if the file mode changed in only one branch, keep the changed
version, and if both branch changed to different mode, mark it as a
conflict.
Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Runtime pack access is done in the pack file mtime order since recent
packs are more likely to contain frequently used objects than old packs.
However the --max-pack-size option can produce multiple packs with mtime
in the reversed order as newer objects are always written first.
Let's modify mtime of later pack files (when any) so they appear older
than preceding ones when a repack creates multiple packs.
Signed-off-by: Nicolas Pitre <nico@cam.org>
The wording of the interactive help text from git-add--interactive.perl is
clearer. Just duplicate that text here.
Signed-off-by: Vineet Kumar <vineet@doorstop.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prior to commit 8320199 (Rewrite builtin-fetch option parsing to use
parse_options().), we understood '-n' as a short option to mean "don't
fetch tags from the remote". This patch reinstates behaviour similar,
but not identical to the pre commit 8320199 times.
Back then, -n always overrode --tags, so if both --tags and -n was
given on command-line, no tags were fetched regardless of argument
ordering. Now we use a "last entry wins" strategy, so '-n --tags'
means "fetch tags".
Since it's patently absurd to say both --tags and --no-tags, this
shouldn't matter in practice.
Spotted-by: Artem Zolochevskiy <azol@altlinux.org>
Reported-by: Dmitry V. Levin <ldv@altlinux.org>
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Tested-by: Andreas Ericsson <ae@op5.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The only reason we did not call "prune" in git-gc was that it is an
inherently dangerous operation: if there is a commit going on, you will
prune loose objects that were just created, and are, in fact, needed by the
commit object just about to be created.
Since it is dangerous, we told users so. That led to many users not even
daring to run it when it was actually safe. Besides, they are users, and
should not have to remember such details as when to call git-gc with
--prune, or to call git-prune directly.
Of course, the consequence was that "git gc --auto" gets triggered much
more often than we would like, since unreferenced loose objects (such as
left-overs from a rebase or a reset --hard) were never pruned.
Alas, git-prune recently learnt the option --expire <minimum-age>, which
makes it a much safer operation. This allows us to call prune from git-gc,
with a grace period of 2 weeks for the unreferenced loose objects (this
value was determined in a discussion on the git list as a safe one).
If you want to override this grace period, just set the config variable
gc.pruneExpire to a different value; an example would be
[gc]
pruneExpire = 6.months.ago
or even "never", if you feel really paranoid.
Note that this new behaviour makes "--prune" be a no-op.
While adding a test to t5304-prune.sh (since it really tests the implicit
call to "prune"), also the original test for "prune --expire" was moved
there from t1410-reflog.sh, where it did not belong.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
commit-tree fails when specifying a remote name (via -r option) and
one of the parent branch has a name. Prefixing with "$remote/" fix it.
Signed-off-by: Marc-Andre Lureau <marcandre.lureau@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
URLs generated by href(..., -replay=>1) (which includes 'next page'
links and alternate view links) didn't set project info correctly
when current page URL is in pathinfo form.
This resulted in broken links such like:
http://www.example.com/w/ARRAY(0x85a5318)?a=shortlog;pg=1
if the 'pathinfo' feature was used, or
http://www.example.com/w/?a=shortlog;pg=1
if it wasn't, instead of correct:
http://www.example.com/w/project.git?a=shortlog;pg=1
This was caused by the fact that href() always replays params in the
arrayref form, were they multivalued or singlevalued, and the code
dealing with 'pathinfo' feature couldn't deal with $params{'project'}
being arrayref.
Setting $params{'project'} is moved before replaying params; this
ensures that 'project' parameter is processed correctly.
Noticed-by: Peter Oberndorfer <kumbayo84@arcor.de>
Noticed-by: Wincent Colaiuta <win@wincent.com>
Signed-off-by: Jakub Narebski <jnareb@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* git://repo.or.cz/git-gui:
git-gui: Simplify MSGFMT setting in Makefile
git-gui: Add option for changing the width of the commit message text box
git-gui: if a background colour is set, set foreground colour as well
git-gui: translate the remaining messages in zh_cn.po to chinese
To prepare msg files for Tcl scripts, the command that is set to MSGFMT
make variable needs to be able to grok "--tcl -l <lang> -d <here>" options
correctly. This patch simplifies the tests done in git-gui's Makefile to
directly test this condition. If the test run does not exit properly with
zero status (either because you do not have "msgfmt" itself, or your
"msgfmt" is too old to grok --tcl option --- the reason does not matter),
have it fall back to po/po2msg.sh
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
* js/remote:
"remote update": print remote name being fetched from
builtin remote rm: remove symbolic refs, too
remote: fix "update [group...]"
remote show: Clean up connection correctly if object fetch wasn't done
builtin-remote: prune remotes correctly that were added with --mirror
Make git-remote a builtin
Test "git remote show" and "git remote prune"
parseopt: add flag to stop on first non option
path-list: add functions to work with unsorted lists
Conflicts:
parse-options.c
* lt/unpack-trees:
unpack_trees(): fix diff-index regression.
traverse_trees_recursive(): propagate merge errors up
unpack_trees(): minor memory leak fix in unused destination index
Make 'unpack_trees()' have a separate source and destination index
Make 'unpack_trees()' take the index to work on as an argument
Add 'const' where appropriate to index handling functions
Fix tree-walking compare_entry() in the presense of --prefix
Move 'unpack_trees()' over to 'traverse_trees()' interface
Make 'traverse_trees()' traverse conflicting DF entries in parallel
Add return value to 'traverse_tree()' callback
Make 'traverse_tree()' use linked structure rather than 'const char *base'
Add 'df_name_compare()' helper function
When the other end has dangling symref, "git fetch" issues an error
message but that is not grave enough to cause the fetch process to fail.
As the result, the user will see something like this:
$ git remote update
error: refs/heads/2.0-uobjects points nowhere!
"remote update" used to report which remote it is fetching from, like
this:
$ git remote update
Updating core
Updating matthieu
error: refs/heads/2.0-uobjects points nowhere!
Updating origin
This reinstates the message "Updating <name>" in "git remote update".
Signed-off-by: Samuel Tardieu <sam@rfc1149.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
git-svn: fix find-rev error message when missing arg
t0021: tr portability fix for Solaris
launch_editor(): allow spaces in the filename
git rebase --abort: always restore the right commit
Just let the user know that a revision argument is missing instead of
a perl error. This error message mimic the "init" error message, but
could be improved.
Signed-off-by: Marc-Andre Lureau <marcandre.lureau@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Solaris' /usr/bin/tr doesn't seem to like multiple character
ranges in brackets (it simply prints "Bad string").
Instead, let's just enumerate the transformation we want.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The construct
sh -c "$0 \"$@\"" <editor> <file>
does not pick up quotes in <editor>, so you cannot give path to the
editor that has a shell IFS whitespace in it, and also give it initial
set of parameters and flags. Replace $0 with <editor> to fix this issue.
This fixes
git config core.editor '"c:/Program Files/What/Ever.exe"'
In other words, you can specify an editor with spaces in its path using a
config containing something like this:
[core]
editor = \"c:/Program Files/Darn/Spaces.exe\"
NOTE: we cannot just replace the $0 with \"$0\", because we still want
this to work:
[core]
editor = emacs -nw
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Previously, --abort would end by git resetting to ORIG_HEAD, but some
commands, such as git reset --hard (which happened in git rebase --skip,
but could just as well be typed by the user), would have already modified
ORIG_HEAD.
Just use the orig-head we store in $dotest instead.
[jc: cherry-picked from 48411d and 4947cf9 on 'master']
Signed-off-by: Mike Hommey <mh@glandium.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Michele Ballabio <barra_cuda@katamail.com> pointed out that gitk
sometimes throws a Tcl error (can't read "yscreen") when switching
views, and proposed a patch. This is a different way of fixing it
which is a bit neater. Basically, in showview we only set yscreen if
the selected commit is on screen to start with, and then we only
scroll the canvas to bring it onscreen if yscreen is set and the
same commit exists in the new view.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Launching gitk on a bare repository or a .git directory
would previously show the work tree as having removed all
files. We now inhibit showing local changes when gitk
is not launched from within a work tree.
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Add test for FREAD_READS_DIRECTORIES to detect when fread() reads fopen'ed
directory.
Tested on these platforms:
AIX 5.3 - FREAD_READS_DIRECTORIES=UnfortunatelyYes
HP-UX B.11.11 - FREAD_READS_DIRECTORIES=UnfortunatelyYes
HP-UX B.11.23 - FREAD_READS_DIRECTORIES=UnfortunatelyYes
Linux 2.6.25-rc4 - FREAD_READS_DIRECTORIES=
Tru64 V5.1 - FREAD_READS_DIRECTORIES=UnfortunatelyYes
Windows - FREAD_READS_DIRECTORIES=
Signed-off-by: Michal Rokos <michal.rokos@nextsoft.cz>
Tested-by: Mike Ralphson <mike@abacus.co.uk>
Tested-by: Johannes Schindelin <Johannes.Schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When skip_unmerged option is not given, unpack_trees() should not just
skip unmerged cache entries but keep them in the result for the caller to
sort them out.
For callers other than diff-index, the incoming index should never be
unmerged, but diff-index is a special case caller.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Being in the project's top directory when starting or continuing a rebase
is not necessary since 533b703 (Allow whole-tree operations to be started
from a subdirectory, 2007-01-12).
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Pressing TAB right after 'git command --long-option=' results in
'git command --long-option=--long-option=' when the long option requires
an argument, but we don't provide completion for its arguments (e.g.
commit --author=, apply --exclude=). This patch detects these long
options and provides empty completion array for them.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
Add the following long options to be completed with command "git":
--paginate
--work-tree=
--help
Signed-off-by: Teemu Likonen <tlikonen@iki.fi>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
When doing completion of rebase options in a subdirectory of the work
tree during an ongoing rebase, wrong options were offered because of the
hardcoded .git/.dotest-merge path.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This patch adds the __git_find_subcommand function, which takes one
argument: a string containing all subcommands separated by spaces. The
function searches through the command line whether a subcommand is
already present. The first found subcommand will be printed to standard
output.
This enables us to remove code duplications from completion functions
for commands having subcommands.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de>
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
There were few places where merge errors detected deeper in the call chain
were ignored and not propagated up the callchain to the caller.
Most notably, this caused switching branches with "git checkout" to ignore
a path modified in a work tree are different between the HEAD version and
the commit being switched to, which it internally notices but ignores it,
resulting in an incorrect two-way merge and loss of the change in the work
tree.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
