Now that we've made the loose SHA1 file reading more careful and
streamlined, we only use the old find_sha1_file() function for checking
whether a loose object file exists at all.
As such, the whole 'return stat information' part of it was just
pointless (nobody cares any more), and the naming of the function is not
really all that relevant either.
So simplify it to not do a 'stat()', but just an existence check (which
is what the callers want), and rename it to 'has_loose_object()' which
matches the use.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
We used to do 'stat()+open()+mmap()+close()' to read the loose object
file data, which does work fine, but has a couple of problems:
- it unnecessarily walks the filename twice (at 'stat()' time and then
again to open it)
- NFS generally has open-close consistency guarantees, which means that
the initial 'stat()' was technically done outside of the normal
consistency rules.
So change it to do 'open()+fstat()+mmap()+close()' instead, which avoids
both these issues.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Avoid cross-directory renames and linking on object creation
Instead of creating new temporary objects in the top-level git object
directory, create them in the same directory they will finally end up in
anyway. This avoids making the final atomic "rename to stable name"
operation be a cross-directory event, which makes it a lot easier for
various filesystems.
Several filesystems do things like change the inode number when moving
files across directories (or refuse to do it entirely).
In particular, it can also cause problems for NFS implementations that
change the filehandle of a file when it moves to a different directory,
like the old user-space NFS server did, and like the Linux knfsd still
does if you don't export your filesystems with 'no_subtree_check' or if
you export a filesystem that doesn't have stable inode numbers across
renames).
This change also obviously implies creating the object fan-out
subdirectory at tempfile creation time, rather than at the final
move_temp_to_file() time. Which actually accounts for most of the size
of the patch.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git add: add long equivalents of '-u' and '-f' options
The option -u stands for --update and it is a good idea to make it clear
especially because this is the only mode of operation of "git add" that
does something different from "adding". Give longer --force synonym to -f
while we are at it as well.
Signed-off-by: SZEDER Gábor <szeder@ira.uka.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-svn: don't append extra newlines at the end of commit messages.
In git, all commits end in exactly one newline character. In svn, commits
end in zero or more newlines. Thus, when importing commits from svn into
git, git-svn always appends two extra newlines to ensure that the
git-svn-id: line is separated from the main commit message by at least one
blank line.
Combined with the terminating newline that's always present in svn commits
produced by git, you usually end up with two blank lines instead of one
between the commit message and git-svn-id: line, which is undesirable.
Instead, let's remove all trailing whitespace from the git commit on the way
through to svn.
Signed-off-by: Avery Pennarun <apenwarr@gmail.com> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-submodule was invoking "die" from within resolve-relative-url, but
this does not actually cause the script to exit. Fix this by returning
the error to the caller and have the caller exit.
While we're at it, clean up the quoting on invocation of
resolve_relative_url as it was wrong.
Signed-off-by: Mark Levedahl <mlevedahl@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-instaweb: improve auto-discovery of httpd and call conventions.
This patch allows calling:
git-instaweb -d apache2
and have the script Do The Right Thing. In particular, the auto-discovery
mechanism has been extended in order to be used for module listing as
well, and the call convention is that if the daemon is apache2/lighttpd
and the parameter to the "-d" option does not end by "-f", the "-f" is
added to the end of the option itself.
Change all backticks to $( ... ) as per Documentation/CodingGuidelines.
Signed-off-by: Flavio Poletti <flavio@polettix.it> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t4126: fix test that happened to work due to timing
The test did "reset --hard" (where the HEAD commit has an empty
blob at path "empty") followed by "> empty", expecting that
the index does not notice the file _changed_ since git wrote
it out upon "reset" if the redirection is done quickly enough.
There was no need to do the emptying, and it gave a wrong result
if "reset --hard" happened on time T and then ">empty" happened on
the next second T+1. This fixes it.
* om/remote-fix:
"remote prune": be quiet when there is nothing to prune
remote show: list tracked remote branches with -n
remote prune: print the list of pruned branches
builtin-remote: split show_or_prune() in two separate functions
remote show: fix the -n option
fast-export: Correctly generate initial commits with no parents
If we are exporting a commit which has no parents we may be doing
it to a branch that already exists, causing fast-import to assume
the branch's current revision should be the sole parent of the
new commit. This can cause `git fast-export | git fast-import`
to produce an incorrect graph for:
A-------M----o------o refs/heads/master
/
B-+
In this graph A and B are initial commits (no parents) but if A was
output first to refs/heads/master and then B is output fast-import
would assume the graph was this instead:
A-------M----o------o refs/heads/master
\ /
+-B-+
Which would cause B, M, and all later commits to have a different
SHA-1, and obviously be quite a different graph.
Sending a reset command prior to B informs fast-import to clear
the implied parent of A, allowing B to remain an initial commit.
Reported-by: Ben Lynn <benlynn@gmail.com> Deemed-obviously-correct-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The behaviour of "sed" on an incomplete line is unspecified by POSIX, and
On Solaris it apparently fails to process input that doesn't end in a LF.
Consequently constructs like
re=$(printf '%s' foo | sed -e 's/bar/BAR/g' $)
cause re to be set to the empty string. Such a construct is used in
git-submodule.sh.
Because the LF at the end of command output are stripped away by the
command substitution, it is a safe and sane change to add a LF at the end
of the printf format specifier.
Signed-off-by: Chris Ridd <chris.ridd@isode.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
user-manual: describe how higher stages are set during a merge
Higher stages store the blobs involved from their side verbatim. Removal
of uninteresting hunks are done by "diff --cc" upon demand and not stored
in the index.
Documentation/git-pull.txt: Use more standard [NOTE] markup
Unlike other manual pages (e.g. git-blame.txt), this used *NOTE:*
to show a side note headed with boldface string "NOTE". Use a paragraph
headed by [NOTE] like others instead.
"remote prune": be quiet when there is nothing to prune
The previous commit made it always say "Pruning $remote" but reported the
URL only when there is something to prune. Make it consistent by not
saying anything at all when there is nothing to prune.
This command is really too quiet which make it unconfortable to use.
Also implement a --dry-run option, in place of the original -n one, to
list stale tracking branches that will be pruned, but do not actually
prune them.
Add a test case for --dry-run.
Signed-off-by: Olivier Marin <dkr@freesurf.fr> Signed-off-by: Junio C Hamano <gitster@pobox.com>
builtin-remote: split show_or_prune() in two separate functions
This allow us to add different features to each of them and keep the
code simple at the same time. Also create a get_remote_ref_states()
to avoid duplicated code.
Signed-off-by: Olivier Marin <dkr@freesurf.fr> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This consolidates the common operations for closing the new temporary file
that we have written, before we move it into place with the final name.
There's some common code there (make it read-only and check for errors on
close), but more importantly, this also gives a single place to add an
fsync_or_die() call if we want to add a safe mode.
This was triggered due to Denis Bueno apparently twice being able to
corrupt his git repository on OS X due to an unlucky combination of kernel
crashes and a not-very-robust filesystem.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
merge-recursive: respect core.autocrlf when writing out the result
The code forgot to convert the blob contents into work tree
representation before writing it out. Also fixes leaks -- earlier
the updated blobs were never freed.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
If you work on a repo with core.autocrlf == true, you would expect
every text file to have CRLF EOLs. However, if you by some operation,
get a conflict, then the conflicted file has LF EOLs.
Now, of course you'd go about resolving the files conflict, and then 'git
add <file>'. When you do that, you'll get the warning saying that LF will
be replaced by CRLF. Then you commit. The end result is that you have a
workingdir with a mix of LF and CRLF files, which after some more
operations may trigger a "whole file changed" diff, due to the workingdir
file now having LF EOLs.
An LF only conflict file results in the resolved file being in LF,
the commit is in LF and a warning saying that LF will be replaced
by CRLF, and the working dir ends up with a mix of CRLF and LF files.
Signed-off-by: Marius Storm-Olsen <marius@trolltech.com> Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Attributes can be specified at three different places: the internal
table of default values, the file $GIT_DIR/info/attributes and files
named .gitattributes in the work tree. Since bare repositories don't
have a work tree, git should ignore any .gitattributes files there.
This patch makes git do that, so the only way left for a user to specify
attributes in a bare repository is the file info/attributes (in addition
to changing the defaults and recompiling).
In addition, git-check-attr is now allowed to run without a work tree.
Like any user of the code in attr.c, it ignores the .gitattributes files
when run in a bare repository. It can still read from info/attributes.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cat-file --batch / --batch-check: do not exit if hashes are missing
Previously, cat-file --batch / --batch-check would silently exit if it
was passed a non-existent SHA1 on stdin. Now it prints "<SHA1>
missing" as in all other cases (and as advertised in the
documentation).
Note that cat-file --batch-check (but not --batch) will still output
"error: unable to find <SHA1>" on stderr if a non-existent SHA1 is
passed, but this does not affect parsing its stdout.
Also, type <= 0 was previously using the potentially uninitialized
type variable (relying on it being 0); it is now being initialized.
Signed-off-by: Lea Wiemann <LeWiemann@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch adds support to compile and run git on 12 additional platforms.
The platforms are based on UNIX Systems Labs (USL)/Novell/SYS V code base.
The most common are Novell UnixWare 2.X.X, SCO UnixWare 7.X.X,
OpenServer 5.0.X, OpenServer 6.0.X, and SCO pre OSR 5 platforms.
Looking at the the various platform headers, I find:
which hides u_short and other typedefs that other header files on these
platforms depend on. WIth _XOPEN_SOURCE defined, sources that include
system header files that depend on the typedefs such as u_short cannot be
compiled on these platforms.
__USLC__ indicates UNIX System Labs Corperation (USLC), or a Novell-derived
compiler and/or some SysV based OS's.
__M_UNIX indicates XENIX/SCO UNIX/OpenServer 5.0.7 and prior releases
of the SCO OS's. It is used just like Apple and BSD, both of these
shouldn't have _XOPEN_SOURCE defined.
This is with suggestions and modifications from
Daniel Barkalow, Junio C Hamano, Thomas Harning, and Jeremy Maitin-Shepard.
Signed-off-by: Boyd Lynn Gerber <gerberb@zenez.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Dynamically sized arrays are gcc and C99 construct. Using them hurts
portability to older compilers, although using them is nice in this case
it is not desirable. This patch removes the only use of the construct
in stop_progress_msg(); the function is about writing out a single line
of a message, and the existing callers of this function feed messages
of only bounded size anyway, so use of dynamic array is simply overkill.
Signed-off-by: Boyd Lynn Gerber <gerberb@zenez.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-send-email: allow whitespace in addressee list
When interactively supplying addresses to send an email to with
send-email, whitespace after the separation comma (as in 'list, jc')
wasn't ignored. This meant that resolving of the alias ' jc' would
fail, sending an email only to list. With this patch, the optional
trailing whitespace is ignored.
Signed-off-by: Pieter de Bie <pdebie@ai.rug.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-rebase -i: mention the short command aliases in the todo list
git rebase -i already supports 'p', 'e' and 's' as aliases for 'pick',
'edit' and 'squash', but one could know it only by reading the source
code. If a user rebases a lot, it's quite handy, so mention these short
forms as well.
Signed-off-by: Miklos Vajna <vmiklos@frugalware.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
An earlier commit aa1dbc9 (Update http-push functionality, 2006-03-07)
borrowed some code from rev-list.c.
This copy and paste made sense back then, because mark_edges_uninteresting(),
and its helper mark_edge_parents_uninteresting(), accessed a file scope
static variable "revs" in rev-list.c, and http-push.c did not have nor care
about such a variable.
But these days they are already properly libified and live in list-objects.c
and they take "revs" as as an argument. Make use of them and lose 20 or
so lines.
git log --graph: print '*' for all commits, including merges
Previously, merge commits were printed with 'M' instead of '*'. This
had the potential to confuse users when not all parents of the merge
commit were included in the log output.
As Junio has pointed out, merge commits can almost always be easily
identified from the log message, anyway.
Signed-off-by: Adam Simpkins <adam@adamsimpkins.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use nonrelative paths instead of absolute paths for cloned repositories
Particularly for the "alternates" file, if one will be created, we
want a path that doesn't depend on the current directory, but we want
to retain any symlinks in the path as given and any in the user's view
of the current directory when the path was given.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This change teaches the graph API that only the first parent of each
commit is interesting when "--first-parent" was specified.
This change also consolidates the graph parent walking logic into two
new internal functions, first_interesting_parent() and
next_interesting_parent(). A simpler fix would have been to simply
break at the end of the 2 existing for loops when
graph->revs->first_parent_only is set. However, this change seems
nicer, especially if we ever need to add any new loops over the parent
list in the future.
Signed-off-by: Adam Simpkins <adam@adamsimpkins.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Print info about "git help COMMAND" on git's main usage pages
Git's main usage pages did not show "git help" as a way to get more
information on a specific subcommand. This patch adds an info line after
the list of git commands currently printed by "git", "git help", "git
--help" and "git help --all".
Signed-off-by: Teemu Likonen <tlikonen@iki.fi> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rewrapped synopsis and removed wrong asterisk behind --count option;
clarified --sort=<key> description for multiple keys; documented that
for-each-ref supports not only glob patterns but also prefixes like
"refs/heads" as patterns, and that multiple patterns can be given.
Signed-off-by: Lea Wiemann <LeWiemann@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
In commit da2478db "describe --always: fall back to showing an
abbreviated object name" we lost the check that skips empty entries in
the object hash table when iterating over it in cmd_name_rev. That may
cause a NULL pointer being handed to show_name(), leading to a
segmentation fault. So add that check back again.
Signed-off-by: Björn Steinbrink <B.Steinbrink@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
gitweb: Add charset info to "raw" output of 'text/plain' blobs
Earlier "blob_plain" view sent "charset=utf-8" only when gitweb
guessed the content type to be text by reading from it, and not when
the MIME type was obtained from /etc/mime.types, or when gitweb
couldn't guess mimetype and used $default_blob_plain_mimetype.
This fixes the bug by always add charset info from
$default_text_plain_charset (if it is defined) to "raw" (a=blob_plain)
output for 'text/plain' blobs.
Generating information for Content-Type: header got separated into
blob_contenttype() subroutine; adding charset info in a special case
was removed from blob_mimetype(), which now should return mimetype
only.
While at it cleanup code a bit: put subroutine parameter
initialization first, make error message more robust (when $file_name
is not defined) if more cryptic, remove unnecessary '"' around
variable ("$var" -> $var).
Signed-off-by: Jakub Narebski <jnareb@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix "git clone $URL" to check out the worktree when asked
The builtin-clone now does the http commit walking and the tree unpacking
in the same process, and the commit walker leaves the in-core objects in a
funny state. When forgetting the data read from the tree object, the
object should be marked "not parsed yet" for later users.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The <pattern> given "git describe --match" was used only to filter tag
objects, and not to filter lightweight tags. This fixes it.
[jc: made the log to clarify this is a bugfix, not an enhancement, with
additional test]
Signed-off-by: Michael Dressel <MichaelTiloDressel@t-online.de> Acked-by: Shawn O. Pearce <spearce@spearce.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: git-log cannot use rev-list specific options
The log family and git-rev-list share the same set of options that come
from revision walking machinery, but they both have options unique to
them. Notably, --header, --timestamp, --stdin and --quiet apply only to
rev-list. Exclude them from the git-log documentation.
Fix t5516 on cygwin: it does not like double slashes at the beginning of a path
The double slashes "//" result from url./$TRASH/. expansion and the
current directory, which even in cygwin contains "/" as first
character. In cygwin such strings have special meaning: UNC path.
Accessing an UNC path built for test purpose usually fails.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
t7502: do not globally unset GIT_COMMITTER_* environment variables
One particular test wants to check the behaviour of the command
when these variables are not set, but the later tests should have
the reliable committer identity for repeatable tests.
Move the "unset" of the variables inside a subshell in the test
that wants to unset them.
cat-file --batch: flush stdout also when objects are missing
cat-file --batch/--batch-check only flushes stdout when the object
exists, but not when it doesn't ("<object> missing"). This makes
bidirectional pipes hang.
Signed-off-by: Lea Wiemann <LeWiemann@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The scripted version of git-commit internally used git-commit-tree which
omitted duplicated parents given from the command line. This prevented a
nonsensical octopus merge from getting created even when you said "git
merge A B" while you are already on branch A.
However, when git-commit was rewritten in C, this sanity check was lost.
This resurrects it.
rebase --interactive: Compute upstream SHA1 before switching branches
If the upstream argument to rebase (the first argument) was relative to
HEAD and the name of the branch to rebase (the second argument) was given,
the upstream would have been interpreted relative to the second argument.
In particular, this command
git rebase -i HEAD topic
would always finish with "Nothing to do". (a1bf91e fixed the same issue
for non-interactive rebase.)
Signed-off-by: Johannes Sixt <johannes.sixt@telecom.at> Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation/git-filter-branch.txt: Fix description of --commit-filter
The old description was misleading and logically impossible. It claimed that
the ancestors of the original commit would be re-written to have the multiple
emitted ids as parents. Not only would this modify existing objects, but it
would create a cycle. What this actually does is pass the multiple emitted ids
to the newly-created children to use as parents.
Signed-off-by: Kevin Ballard <kevin@sb.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix search form generation to not modify $cgi->param(...)'s.
In git_header_html() we used to use $cgi->hidden(-name => "a") etc. to
generate hidden fields; unfortunately to use this form it is required
to modify $cgi->param("a") etc., which makes href(-replay,...) use
wrong replay values. This for example made the "next" link on the
bottom of the page has a=search instead of a=$action, and thus fails to
get you to the next page.
Because in CGI the value of a hidden field is "sticky", there is no
way to modify it short of modifying $cgi->param(...). Therefore it
got replaced by generating <input type="hidden" ...> element [semi]
directly.
Alternate solution would be for href(-replay,...) to use values saved
in global variables, such as $action etc., instead of (re)reading them
from $cgi->param($symbol).
The bad link was reported by Kai Blin through
http://bugs.debian.org/481902
Reported-by: Kai Blin <kai.blin@gmail.com> Signed-off-by: Jakub Narebski <jnareb@gmail.com> Tested-by: Gerrit Pape <pape@smarden.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Revert "git.el: Set process-environment instead of invoking env"
This reverts commit dbe48256b41c1e94d81f2458d7e84b1fdcb47026, which
caused mis-encoding of non-ASCII author/committer names when the
git-status mode is used to create commits.
Signed-off-by: Karl Hasselström <kha@treskal.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* sp/remote:
Make "git-remote rm" delete refs acccording to fetch specs
Make "git-remote prune" delete refs according to fetch specs
Remove unused remote_prefix member in builtin-remote