948dd34 (diff-index: careful when inspecting work tree items, 2008-03-30)
made the work tree check careful not to be fooled by a new directory that
exists at a place the index expects a blob. For such a change to be a
typechange from blob to submodule, the new directory has to be a
repository.
However, if the index expects a submodule there, we should not insist the
work tree entity to be a repository --- a simple directory that is not a
full fledged repository (even an empty directory would do) should be
considered an unmodified subproject, because that is how a superproject
with a submodule is checked out sparsely by default.
This makes the function check_work_tree_entity() even more careful not to
report a submodule that is not checked out as removed. It fixes the
recently added test in t4027.
checkout: don't rfc2047-encode oneline on detached HEAD
When calling pretty_print_commit, there is an implicit
assumption that passing in a non-NULL "subject" variable
for oneline or email formats means that the output is part
of a subject and therefore "subject" to rfc2047 encoding.
This is not the desired effect when reporting the movement
of detached HEAD.
When the delta data is cached in memory until it is written to a pack
file on disk, it is best to compress it right away in find_deltas() for
the following reasons:
- we have to compress that data anyway;
- this allows for caching more deltas with the same cache size limit;
- compression is potentially threaded.
This last point is especially relevant for SMP run time. For example,
repacking the Linux repo on a quad core processor using 4 threads with
all default settings produce the following results before this change:
real 2m27.929s
user 4m36.492s
sys 0m3.091s
And with this change applied:
real 2m13.787s
user 4m37.486s
sys 0m3.159s
So the actual execution time stayed more or less the same but the
wall clock time is shorter.
This is however not a good thing to do when generating a pack for
network transmission. In that case, the network is most likely to
throttle the data throughput, so it is best to make find_deltas()
faster in order to start writing data ASAP since we can afford
spending more time between writes to compress the data
at that point.
Signed-off-by: Nicolas Pitre <nico@cam.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
diff: make "too many files" rename warning optional
In many cases, the warning ends up as clutter, because the
diff is being done "behind the scenes" from the user (e.g.,
when generating a commit diffstat), and whether we show
renames or not is not particularly interesting to the user.
However, in the case of a merge (which is what motivated the
warning in the first place), it is a useful hint as to why a
merge with renames might have failed.
This patch makes the warning optional based on the code
calling into diffcore. We default to not showing the
warning, but turn it on for merges.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current rename limit default of 100 was arbitrarily
chosen. Testing[1] has shown that on modern hardware, a
limit of 200 adds about a second of computation time, and a
limit of 500 adds about 5 seconds of computation time.
This patch bumps the default limit to 200 for viewing diffs,
and to 500 for performing a merge. The limit for generating
git-status templates is set independently; we bump it up to
200 here, as well, to match the diff limit.
[1]: See <20080211113516.GB6344@coredump.intra.peff.net>
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of rename limiting is to bound the amount of time
we spend figuring out inexact renames. Currently we use a
single value, diff.renamelimit, for all situations. However,
it is probably the case that a user is willing to spend more
time finding renames during a merge than they are while
looking at git-log.
This patch provides a way of setting those values separately
(though for backwards compatibility, merge still falls back
on the diff renamelimit).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
log: print log entry terminator even if the message is empty
This eliminates a special case in the show_log() function, to help
simplify the terminator semantics. Now show_log() always prints a
newline after the log entry when use_terminator is set, even if the log
message is empty.
This change should only affect the --pretty=tformat output, since that
was the only way to trigger this special case.
Signed-off-by: Adam Simpkins <adam@adamsimpkins.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Disable "Reset %s branch to here" when on a detached head
When we are on a detached head - since gitk does not display where
we are - reset has no sense, so disable the relevant line on the
context menu, and point out to the user that we are on a detached head.
Otherwise, a reset from gitk when on a detached head returns the
error:
can't read "headids()": no such element in array
can't read "headids()": no such element in array
while executing
"removehead $headids($name) $name"
(procedure "movehead" line 4)
invoked from within
"movehead $newhead $mainhead"
(procedure "readresetstat" line 20)
invoked from within
"readresetstat file4"
("eval" body line 1)
invoked from within
"eval $script"
(procedure "dorunq" line 9)
invoked from within
"dorunq"
("after" script)
[paulus@samba.org: changed menu item to "Detached head: can't reset"]
Signed-off-by: Michele Ballabio <barra_cuda@katamail.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
gitk: Makefile/install: force permissions when installing files and dirs
The msg-files msgs/*.msg used to be installed with mode 755 although
they're not executables. With this commit, files are forced to be
installed with mode 644, directories and executables with mode 755.
I copied the Italian translation and translated the strings
to Spanish starting from there. This incorporates suggestions
from Wincent Colaiuta and Carlos Rica.
Signed-off-by: Santiago Gala <sgala@apache.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
When using mixes of imports, from CVS and SVN into the same git
repository, you'd like to harmonise the imports to the format cvsimport
uses.
git-svn supports an experimental option --use-log-author which currently
results in the same logentry as without that option when no From: or Signed-off-by: is found in the logentry ($email currently ends up empty,
and hence is generated again).
This patches harmonises the result with cvsimport, and makes
git-svn --use-log-author produce:
Author: foobaruser <foobaruser>
Signed-off-by: Stephen R. van den Berg <srb@cuci.nl> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Most users should be using git-gc instead of directly
calling prune. For those who really do want more information
on pruning, let's point them at git-fsck, which goes into
slightly more detail on reachability.
And since we're pointing users there, let's make sure
reflogs are mentioned in git-fsck(1).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-svn: detect and fail gracefully when dcommitting to a void
The command
git svn clone (URL of an empty SVN repo here)
works, creates an empty git repository. I can perform the initial
commit there, but then, "git svn dcommit" says :
Use of uninitialized value in concatenation (.) or string at .../git-svn line 414.
Committing to ...
Unable to determine upstream SVN information from HEAD history
I guess a correct management of the initial commit in git-svn would be
hard to implement, but at least, the error message can be improved.
First step is something like the patch below, and better would be for
"git svn clone" to warn that it won't be able to do much with the
cloned repo.
Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
cvsimport: always pass user data to "system" as a list
fix reflog approxidate parsing bug
Fix use after free() in builtin-fetch
fetch-pack: do not stop traversing an already parsed commit
Use "=" instead of "==" in condition as it is more portable
into its constituent ref, reflog date, and path components.
We never actually munge the string itself, but instead keep
offsets into the string with their associated lengths.
When we call approxidate on the contents inside braces,
however, we pass just a string without a length. This means
that approxidate could sometimes look past the closing brace
and (erroneously) interpret the rest of the string as part
of the date.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The purpose of --first-parent is to view the tree without looking at
side branche. This is accomplished by pretending there are no other
parents than the first parent when encountering a merge.
The current code marks the other parents as seen, which means that the tree
traversal will behave differently depending on the order merges are handled.
When a fast forward is artificially recorded as a merge,
-----
/ \
D---E---F---G master
the current first-parent code considers E to be seen and stops the
traversal after showing G and F.
Signed-off-by: Stephen R. van den Berg <srb@cuci.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
(gdb) bt
0 0x000000349fd6d44b in free () from /lib64/libc.so.6
1 0x000000000048f4eb in transport_unlock_pack (transport=0x7ce530) at transport.c:811
2 0x000000349fd31b25 in exit () from /lib64/libc.so.6
3 0x00000000004043d8 in handle_internal_command (argc=3, argv=0x7fffea4449f0) at git.c:379
4 0x0000000000404547 in main (argc=3, argv=0x7fffea4449f0) at git.c:443
5 0x000000349fd1c784 in __libc_start_main () from /lib64/libc.so.6
6 0x0000000000403ef9 in ?? ()
7 0x00007fffea4449d8 in ?? ()
8 0x0000000000000000 in ?? ()
I then remembered, my .bashrc has this..
export MALLOC_PERTURB_=$(($RANDOM % 255 + 1))
which is handy for showing up such bugs.
More info on this glibc feature is at http://udrepper.livejournal.com/11429.html
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
fetch-pack: do not stop traversing an already parsed commit
f3ec549 (fetch-pack: check parse_commit/object results, 2008-03-03)
broke common ancestor computation by stopping traversal when it sees
an already parsed commit. This should fix it.
Option is only completed when .git/MERGE_HEAD is present.
Signed-off-by: Richard Quirk <richard.quirk@gmail.com> Acked-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Fix handling of tree file list with special chars in names
Alex Riesen pointed out that displaying a commit in 'tree' mode fails
if some files have names with special characters such as '{' or '}' in
them, due to the fact that we treat the line returned from git ls-tree
as a Tcl list at one point.
This fixes it by doing what I originally intended but didn't quite
get right. We split the line from git ls-tree at the first tab and
treat the part before the tab as a list (which is OK since it doesn't
have special characters in it) and the part after the tab as the
filename.
Michael G. Noll said in comments to the "Switching my code repository from
Subversion (SVN) to git" article (http://tinyurl.com/37v67l) in his "My
digital moleskine" blog, that one of the things he is missing in gitweb
from SVN::Web is an RSS feed with news/information of the current view
(including RSS feed for single file or directory).
This is not exactly true, as since refactoring feed generation in af6feeb
(gitweb: Refactor feed generation, make output prettier, add Atom feed,
2006-11-19), gitweb can generate feeds (RSS or Atom) for history of a
given branch, history limited to a given directory, or history of a given
file. Nevertheless this required handcrafting the URL to get wanted RSS
feed.
This commit makes gitweb select feed links in the HTML header and in
page footer depending on current view (action). It is more elaborate,
and I guess more correct, than simple patch adding $hash ('h')
parameter to *all* URLs, including feed links, by Jean-Baptiste Quenot
Subject: [PATCH] gitweb: Add hash parameter in feed URL when a hash
is specified in the current request
Message-ID: <ae63f8b50803211138y6355fd11pa64cda50a1f53011@mail.gmail.com>
If $hash ('h') or $hash_base ('hb') parameter is a branch name
(i.e. it starts with 'refs/heads/'; all generated URLs use this form
to discriminate between tags and heads), it is used in feed URLs; if
$file_name ('f') is defined, it is used in feed URLs. Feed title is
set according to the kind of web feed: it is either 'log' for generic
feed, 'log of <branch>', 'history of <filename>' for generic history
(using implicit or explicit HEAD, i.e. current branch) or 'history of
<filename> on <branch>'.
There are special cases: 'heads' and 'forks' views should use OPML
providing list of available feeds; 'tags' probably also should use
OPML; there is no web feed equivalent to 'search' view. Currently all
those cases fallback to (show) default feed. Such feed link uses
"generic" class, and is shown in slightly lighter color for
distinction.
Currently feed can have but one starting point, and does not support
negative (exclude) commit arguments. Therefore for now for *diff
views it is chosen that feed follow the "to" part: to-name, to-commit
for 'blobdiff', 'treediff' and 'commitdiff' views.
Generating parameters for href() for feed link was separated
(refactored) into get_feed_info() subroutine.
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git add *" is actually fundamentally different from "git add .", and
yeah, you should generally use the latter.
The reason? The argument list is actually something different from what
you think it is. For git, it's a "pathspec", so what actualy happens is
that in *both* cases, it will really traverse the whole tree, and then
match every file it finds against the pathspec.
So think of the arguments not as a file list, but as a random bunch of
patterns to match against the files you have!
Which is why the cost is actually approximately O(n*m), where "n" is the
size of the working tree, and "m" is the number of pathspecs.
So the reason "git add ." is fast is actually that "m" in that case is
just 1 (just one trivial pattern), and then "git add *" is slow because
"m" is large (lots of complicated patterns). In both cases, 'n' is the
same (== the whole set of files in your working tree).
Anyway, here's a trivial patch that doesn't change this fundamental fact,
but that avoids doing anything *expensive* until we've done some cheap
initial tests. It may or may not help your test-case, but it's pretty
simple and it matches the other git optimizations in this area (ie
"conceptually handle the general case, but optimize the simple cases where
we can exit early")
Notice how this patch doesn' actually change the fundamental O(n^2)
behaviour, but it makes it much cheaper by generally avoiding the
expensive 'fnmatch' and 'strlen/strncmp' when they are obviously not
needed.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This makes a struct ref able to represent a symref, and makes http.c
able to recognize one, and makes transport.c look for "HEAD" as a ref
in the list, and makes it dereference symrefs for the resulting ref,
if any.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This simplifies a few things, makes a few things slightly more
complicated, but, more importantly, allows that, when struct ref can
represent a symref, http_fetch_ref() can return one.
Incidentally makes the string that http_fetch_ref() gets include "refs/"
(if appropriate), because that's how the name field of struct ref works.
As far as I can tell, the usage in walker:interpret_target() wouldn't have
worked previously, if it ever would have been used, which it wouldn't
(since the fetch process uses the hash instead of the name of the ref
there).
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
help: use "man.<tool>.cmd" as custom man viewer command
Currently "git help -m GITCMD" is restricted to a set of man viewers
defined at compile time. You can subvert the "man.<tool>.path" to
force "git help -m" to use a different man, viewer, but if you have a
man viewer 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 "man.<tool>.cmd" which allows a
more flexible man viewer choice.
If you run "git help -m GITCMD" with the "man.viewer" config variable
set to an unrecognized tool then it will query the "man.<tool>.cmd"
config variable. If this variable exists, then the specified tool will
be treated as a custom man viewer and it will be run in a shell with
the man page name of the GITCMD added as extra parameter.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
help: use man viewer path from "man.<tool>.path" config var
This patch implements reading values from "man.<tool>.path"
configuration variables, and using these values as pathes to
the man viewer <tool>s when lauching them.
This makes it possible to use different version of the tools
than the one on the current PATH, or maybe a custom script.
In this patch we also try to launch "konqueror" using
"kfmclient" even if a path to a konqueror binary is given
in "man.konqueror.path".
The "man_viewer_list" becomes a simple string list to simplify
things for the following patches.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allow cherry-pick (and revert) to add signoff line
I often find myself pulling patches off of other peoples trees using
cherry-pick, and following it with an immediate 'git commit --amend -s'
command. Eliminate the need for a double commit by allowing signoff on a
cherry-pick or revert.
Signed-off-by: Dan McGee <dpmcgee@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitk: Reorganize processing of arguments for git log
This moves the scanning of the argument list for each view into a
new function, parseviewargs, which is called from start_rev_list.
This also makes the date mode and the merge mode be per-view rather
than global. In merge mode, we work out the list of relevant files
in a new function called from start_rev_list, so it will be updated
on File->Reload. Plus we now do that after running the argscmd, so
if we have one and it generates a -d or --merge option they will be
correctly handled now.
The other thing this does is to make errors detected in start_rev_list
not be fatal. Now instead of doing exit 1 we just pop up and error
window and put "No commits selected" in the graph pane.
* maint:
remote: create fetch config lines with '+'
push: allow unqualified dest refspecs to DWIM
doc/git-gc: add a note about what is collected
t5516: remove ambiguity test (1)
Linked glossary from cvs-migration page
write-tree: properly detect failure to write tree objects
1. check for an unambiguous 'dst' on the remote; if it
exists, then push to that ref
2. otherwise, check if 'dst' begins with 'refs/'; if it
does, create a new ref
3. otherwise, complain because we don't know where in the
refs hierarchy to put 'dst'
However, in some cases, we can guess about the ref type of
'dst' based on the ref type of 'src'. Specifically, before
complaining we now check:
2.5. if 'src' resolves to a ref starting with refs/heads
or refs/tags, then prepend that to 'dst'
So now this creates a new branch on the remote, whereas it
previously failed with an error message:
git push master:newbranch
Note that, by design, we limit this DWIM behavior only to
source refs which resolve exactly (including symrefs which
resolve to existing refs). We still complain on a partial
destination refspec if the source is a raw sha1, or a ref
expression such as 'master~10'.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
It seems to be a FAQ that people try running git-gc, and
then get puzzled about why the size of their .git directory
didn't change. This note mentions the reasons why things
might unexpectedly get kept.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-gui: Don't use '$$cr master' with aspell earlier than 0.60
Apparently aspell 0.50 does not recognize "$$cr master" as a command,
but instead tries to offer suggestions for how to correctly spell
the word "cr". This is not quite what we are after when we want
the name of the current dictionary.
Instead of locking up git-gui waiting for a response that may never
come back from aspell we avoid sending this command if the binary
we have started claims to be before version 0.60.
Signed-off-by: Shawn O. Pearce <spearce@spearce.org>
This test tried to push into a remote with ambiguous refs in
remotes/$x/master and remotes/$y/master. However, the remote
never actually tells us about the refs/remotes hierarchy, so
we don't even see this ambiguity.
The test happened to pass because we were simply looking for
failure, and the test fails for another reason: the dst
refspec does not exist and does not begin with refs/, making
it invalid.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Coming from CVS, I found the git glossary vital to learning git and learning
how terms in git correlate to the cvs terminology with which I am familiar.
This patch links the glossary from the cvs-migration page so cvs users will
be able to fine the glossary as soon as they start looking at git documents.
Signed-off-by: Matt Graham <mdg149@gmail.com> Acked-by: J. Bruce Fields <bfields@fieldses.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
write-tree: properly detect failure to write tree objects
Tomasz Fortuna reported that "git commit" does not error out properly when
it cannot write tree objects out. "git write-tree" shares the same issue,
as the failure to notice the error is deep in the logic to write tree
objects out recursively.
* maint:
Amend git-push refspec documentation
git-gc --prune is deprecated
svn-git: Use binmode for reading/writing binary rev maps
diff options documentation: refer to --diff-filter in --name-status
Don't force imap.host to be set when imap.tunnel is set
git-clone.txt: Adjust note to --shared for new pruning behavior of git-gc
git-svn bug with blank commits and author file
archive.c: format_subst - fixed bogus argument to memchr
copy.c: copy_fd - correctly report write errors
gitattributes: Fix subdirectory attributes specified from root directory
25ee9731c137d0a24b0f4879eb0b0cce9b77d5b0 made the '--prune' option
deprecated and removed its description from the git-gc man page. This
patch removes all references to this option from the rest of the Git
documentation.
svn-git: Use binmode for reading/writing binary rev maps
Otherwise, there is a possible interaction with UTF-8 locales in
combination with PERL_UNICODE, resulting in "inconsistent size: 40" or
"read:"-type errors.
See also:
perldoc -f binmode
<http://perldoc.perl.org/perl581delta.html#UTF-8-no-longer-default-under-UTF-8-locales>
Signed-off-by: Michael Weber <michaelw@foldr.org> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
which works wonderfully and doesn't require a username, password or port
setting.
However, git-imap-send currently requires that the imap.host variable be
set in the config even when it was unused. This patch changes imap-send
to only require that the imap.host setting is set if imap.tunnel is not
set. Otherwise, server.host is set to "tunnel" for reporting purposes.
Acked-by: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-clone.txt: Adjust note to --shared for new pruning behavior of git-gc
Since git-gc now always calls prune, even with --auto, unreferenced objects
may be removed by more operations than just git-gc. This is important for
clones created using --shared or --reference.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil> Signed-off-by: Dmitry Potapov <dpotapov@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
When trying to import from svn using an author file, git-svn bails out
if it encounters a blank author. The attached patch changes this
behavior and allow using the author file with blanks authors.
I came across this bug while importing from a cvs2svn repo where the
initial revision (1) has a blank author. This doesn't break the behavior
of bailing out when an unknown author is encountered.
Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* maint:
post-receive-email: fix accidental removal of a trailing space in signature line
Escape project names before creating pathinfo URLs
Escape project name in regexp
bash: Add completion for git diff --base --ours --theirs
diff-options.txt: document the new "--dirstat" option
post-receive-email: fix accidental removal of a trailing space in signature line
post-receive-email adds a signature to the end of emails in
generate_email_footer(). The signature was separated from the main email
body using the standard string "-- ". (see RFC 3676)
a6080a0 (War on whitespace, 2007-06-07) removed the trailing whitespace
from "-- ", leaving it as "--", which is not a correct signature
separator.
This patch restores the missing space, but does it in a way that will
not set off the trailing whitespace alarms.
Signed-off-by: Andy Parkins <andyparkins@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>