After "git add -N" and then writing a tree object out of the
index, the cache-tree data structure got corrupted.
* nd/invalidate-i-t-a-cache-tree:
cache-tree: invalidate i-t-a paths after generating trees
cache-tree: fix writing cache-tree when CE_REMOVE is present
cache-tree: replace "for" loops in update_one with "while" loops
cache-tree: remove dead i-t-a code in verify_cache()
The way "git svn" asked for password using SSH_ASKPASS and
GIT_ASKPASS was not in line with the rest of the system.
* ss/svn-prompt:
git-svn, perl/Git.pm: extend and use Git->prompt method for querying users
perl/Git.pm: Honor SSH_ASKPASS as fallback if GIT_ASKPASS is not set
git-svn, perl/Git.pm: add central method for prompting passwords
t9200 runs "cvs init" on a directory that already exists, but a
platform can configure this fail for the current user (e.g. you need
to be in the cvsadmin group on NetBSD 6.0).
* jc/test-cvs-no-init-in-existing-dir:
t9200: let "cvs init" create the test repository
An element on GIT_CEILING_DIRECTORIES list that does not name the
real path to a directory (i.e. a symbolic link) could have caused
the GIT_DIR discovery logic to escape the ceiling.
* mh/ceiling:
string_list_longest_prefix(): remove function
setup_git_directory_gently_1(): resolve symlinks in ceiling paths
longest_ancestor_length(): require prefix list entries to be normalized
longest_ancestor_length(): take a string_list argument for prefixes
longest_ancestor_length(): use string_list_split()
Introduce new function real_path_if_valid()
real_path_internal(): add comment explaining use of cwd
Introduce new static function real_path_internal()
* sl/maint-git-svn-docs:
git-svn: Note about tags.
git-svn: Expand documentation for --follow-parent
git-svn: Recommend use of structure options.
git-svn: Document branches with at-sign(@).
Document that 'git svn' will import SVN tags as branches.
Signed-off-by: Sebastian Leske <sebastian.leske@sleske.name> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Describe what the option --follow-parent does, and what happens if it is
set or unset.
Signed-off-by: Sebastian Leske <sebastian.leske@sleske.name> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Document that when using git svn, one should usually either use the
directory structure options to import branches as branches, or only
import one subdirectory. The default behaviour of cloning all branches
and tags as subdirectories in the working copy is usually not what the
user wants.
Signed-off-by: Sebastian Leske <sebastian.leske@sleske.name> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git svn sometimes creates branches with an at-sign in the name
(branchname@revision). These branches confuse many users and it is a FAQ
why they are created. Document when git svn creates them.
Signed-off-by: Sebastian Leske <sebastian.leske@sleske.name> Acked-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-remote-helpers.txt: document invocation before input format
In the distant past, the order things were documented was
'Invocation', 'Commands', 'Capabilities', ...
Then it was decided that before giving a list of Commands, there
should be an overall description of the 'Input format', which was
a wise decision. However, this description was put as the very
first thing, with the rationale that any implementor would want
to know that first.
However, it seems an implementor would actually first need to
know how the remote helper will be invoked, so moving
'Invocation' to the front again seems logical. Moreover, we now
don't switch from discussing the input format to the invocation
style and then back to input related stuff.
Signed-off-by: Max Horn <max@quendi.de> Acked-by: Sverre Rabbelier <srabbelier@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: move diff.wordRegex from config.txt to diff-config.txt
19299a8 (Documentation: Move diff.<driver>.* from config.txt to
diff-config.txt, 2011-04-07) moved the diff configuration options to
diff-config.txt, but forgot about diff.wordRegex, which was left
behind in config.txt. Fix this.
Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The optional third parameter when __git_ps1 is used in
PROMPT_COMMAND mode as format string for printf to further
customize the way the git status string is embedded in the
user's PS1 prompt.
Signed-off-by: Simon Oosthoek <s.oosthoek@xs4all.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jk/mailmap-cleanup:
contrib: update stats/mailmap script
.mailmap: normalize emails for Linus Torvalds
.mailmap: normalize emails for Jeff King
.mailmap: fix broken entry for Martin Langhoff
.mailmap: match up some obvious names/emails
* ta/doc-cleanup:
Documentation: build html for all files in technical and howto
Documentation/howto: convert plain text files to asciidoc
Documentation/technical: convert plain text files to asciidoc
Change headline of technical/send-pack-pipeline.txt to not confuse its content with content from git-send-pack.txt
Shorten two over-long lines in git-bisect-lk2009.txt by abbreviating some sha1
Split over-long synopsis in git-fetch-pack.txt into several lines
"git diff <blob> <blob>" was not documented and was only hinted as
an extension to "git diff <commit> <commit> -- <pathspec>", but
comparison between two blobs are more special than that. It does
not take any pathspec to begin with.
* sl/git-svn-docs:
git-svn: Note about tags.
git-svn: Expand documentation for --follow-parent
git-svn: Recommend use of structure options.
git-svn: Document branches with at-sign(@).
* jk/mailmap-cleanup:
contrib: update stats/mailmap script
.mailmap: normalize emails for Linus Torvalds
.mailmap: normalize emails for Jeff King
.mailmap: fix broken entry for Martin Langhoff
.mailmap: match up some obvious names/emails
Move ./technical/api-command.txt to ./howto/new-command.txt
The contents of this document does not describe any particular API, but
is more about the way to add a new command, which belongs to the "How To"
section of the documentation suite.
Signed-off-by: Thomas Ackermann <th.acker@arcor.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
http.c: Avoid username prompt for certifcate credentials
If sslCertPasswordProtected is set to true do not ask for username to decrypt rsa key. This question is pointless, the key is only protected by a password. Internaly the username is simply set to "".
Signed-off-by: Rene Bredlau <git@unrelated.de> Acked-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
refs: do not use cached refs in repack_without_ref
When we delete a ref that is packed, we rewrite the whole
packed-refs file and simply omit the ref that no longer
exists. However, we base the rewrite on whatever happens to
be in our refs cache, not what is necessarily on disk. That
opens us up to a race condition if another process is
simultaneously packing the refs, as we will overwrite their
newly-made pack-refs file with our potentially stale data,
losing commits.
# in one terminal, repack the refs repeatedly
cd parent &&
while true; do
git pack-refs --all
done
# in another terminal, simultaneously push updates to
# master, and create and delete an unrelated ref
cd child &&
while true; do
git push origin HEAD:newbranch &&
git commit --allow-empty -m foo
us=`git rev-parse master` &&
git push origin master &&
git push origin :newbranch &&
them=`git --git-dir=../parent rev-parse master` &&
if test "$them" != "$us"; then
echo >&2 "$them" != "$us"
exit 1
fi
done
In many cases the two processes will conflict over locking
the packed-refs file, and the deletion of newbranch will
simply fail. But eventually you will hit the race, which
happens like this:
1. We push a new commit to master. It is already packed
(from the looping pack-refs call). We write the new
value (let us call it B) to $GIT_DIR/refs/heads/master,
but the old value (call it A) remains in the
packed-refs file.
2. We push the deletion of newbranch, spawning a
receive-pack process. Receive-pack advertises all refs
to the client, causing it to iterate over each ref; it
caches the packed refs in memory, which points at the
stale value A.
3. Meanwhile, a separate pack-refs process is running. It
runs to completion, updating the packed-refs file to
point master at B, and deleting $GIT_DIR/refs/heads/master
which also pointed at B.
4. Back in the receive-pack process, we get the
instruction to delete :newbranch. We take a lock on
packed-refs (which works, as the other pack-refs
process has already finished). We then rewrite the
contents using the cached refs, which contain the stale
value A.
The resulting packed-refs file points master once again at
A. The loose ref which would override it to point at B was
deleted (rightfully) in step 3. As a result, master now
points at A. The only trace that B ever existed in the
parent is in the reflog: the final entry will show master
moving from A to B, even though the ref still points at A
(so you can detect this race after the fact, because the
next reflog entry will move from A to C).
We can fix this by invalidating the packed-refs cache after
we have taken the lock. This means that we will re-read the
packed-refs file, and since we have the lock, we will be
sure that what we read will be atomically up-to-date when we
write (it may be out of date with respect to loose refs, but
that is OK, as loose refs take precedence).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
These "expect-failure" tests were not looking for the right string
in the patch file. For example:
grep "^ *"S. E. Cipient" <scipient@example.com>\$" patch5
was looking for "^ *S." in these three files:
"E."
"Cipient <scipient@example.com>$"
"patch5"
With some implementations of grep, the lack of file "E." was
reported as an error, leading to the failure of the test.
With other implementations of grep, the pattern "^ *S." matched what
was in patch5, without diagnosing the missing files as an error, and
made these tests unexpectedly pass.
The check_snapshot function makes sure that no cruft outside the
repository hierarchy is added to the tar archive. The output from
"tar tf" on the resulting archive is inspected to see if there is
anything that does not begin with "$prefix/".
There are two issues with this implementation:
- Traditional tar implemenations that do not understand
pax_global_header will write it out as if it is a plain file at
the top-level;
- Some implementations of tar do not add trailing slash when
showing a directory entry (i.e. the output line for the entire
archive will show "$prefix", not "$prefix/").
Fix them so that what we want to validate can be tested with
traditional tar implementations.
On systems without "locale" installed, t0200-gettext-basic.sh leaked
error messages when checking if some test locales are available.
Hide them, as they are not very useful.
xmkstemp(): avoid showing truncated template more carefully
Some implementations of xmkstemp() leaves the given in/out buffer
truncated when they return with failure.
6cf6bb3 (Improve error messages when temporary file creation fails,
2010-12-18) attempted to show the real filename we tried to create
(but failed), and if that is not available due to such truncation,
to show the original template that was given by the caller.
But it failed to take into account that the given template could
have "directory/" in front, in which case the truncation point may
not be template[0] but somewhere else.
As it was not a common operation, it was described as if it is a
side note for the more common two-commit variant, but this mode
behaves very differently, e.g. it does not make any sense to ask
recursive behaviour, or give the command a pathspec.
t7004: do not create unneeded gpghome/gpg.conf when GPG is not used
These tests themselves are properly protected by the GPG
prerequisite, but one of the set-up steps outside the
test_expect_success block unconditionally assumed that there is a
gpghome/ directory, which is not true if GPG is not being used.
It may be a good idea to move the whole set-up steps in the test but
that is a follow-up topic.
The forms of checkout that do not take a path are lumped together in
the DESCRIPTION section, but the description for this group is
dominated by explanation of the -b|-B form.
Split these apart for more clarity.
Signed-off-by: Chris Rorvick <chris@rorvick.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git log -p -S<string>" now looks for the <string> after applying
the textconv filter (if defined); earlier it inspected the contents
of the blobs without filtering.
git-svn, perl/Git.pm: extend and use Git->prompt method for querying users
git-svn reads usernames and other user queries from an interactive
terminal. This cause GUIs (w/o STDIN connected) to hang waiting forever
for git-svn to complete (http://code.google.com/p/tortoisegit/issues/detail?id=967).
This change extends the Git::prompt helper, so that it can also be used
for non password queries, and makes use of it instead of using
hand-rolled prompt-response code that only works with the interactive
terminal.
Signed-off-by: Sven Strickroth <email@cs-ware.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-svn, perl/Git.pm: add central method for prompting passwords
git-svn reads passwords from an interactive terminal or by using
GIT_ASKPASS helper tool. This cause GUIs (w/o STDIN connected) to hang
waiting forever for git-svn to complete
(http://code.google.com/p/tortoisegit/issues/detail?id=967).
Instead of using hand-rolled prompt-response code that only works with the
interactive terminal, a reusable prompt() method is introduced in this commit.
Signed-off-by: Sven Strickroth <email@cs-ware.de> Signed-off-by: Junio C Hamano <gitster@pobox.com>
SubmittingPatches: add convention of prefixing commit messages
Conscientious newcomers to git development will read SubmittingPatches
and CodingGuidelines, but could easily miss the convention of
prefixing commit messages with a single word identifying the file
or area the commit touches.
Signed-off-by: Adam Spiers <git@adamspiers.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Documentation: don't link to example mail addresses
Email addresses in documentation are converted into mailto: hyperlinks
in the HTML output and footnotes in man pages. This isn't desirable for
cases where the address is used as an example and is not valid.
Particularly annoying is the example "jane@laptop.(none)" which appears
in git-shortlog(1) as "jane@laptop[1].(none)", with note 1 saying:
1. jane@laptop
mailto:jane@laptop
Fix this by escaping these email addresses with a leading backslash, to
prevent Asciidoc expanding them as inline macros.
In the case of mailmap.txt, render the address monospaced so that it
matches the block examples surrounding that paragraph.
Helped-by: Jeff King <peff@peff.net> Signed-off-by: John Keeping <john@keeping.me.uk> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cache-tree: invalidate i-t-a paths after generating trees
Intent-to-add entries used to forbid writing trees so it was not a
problem. After commit 3f6d56d (commit: ignore intent-to-add entries
instead of refusing - 2012-02-07), we can generate trees from an index
with i-t-a entries.
However, the commit forgets to invalidate all paths leading to i-t-a
entries. With fully valid cache-tree (e.g. after commit or
write-tree), diff operations may prefer cache-tree to index and not
see i-t-a entries in the index, because cache-tree does not have them.
Reported-by: Jonathon Mah <me@JonathonMah.com> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cache-tree: fix writing cache-tree when CE_REMOVE is present
entry_count is used in update_one() for two purposes:
1. to skip through the number of processed entries in in-memory index
2. to record the number of entries this cache-tree covers on disk
Unfortunately when CE_REMOVE is present these numbers are not the same
because CE_REMOVE entries are automatically removed before writing to
disk but entry_count is not adjusted and still counts CE_REMOVE
entries.
Separate the two use cases into two different variables. #1 is taken
care by the new field count in struct cache_tree_sub and entry_count
is prepared for #2.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cache-tree: replace "for" loops in update_one with "while" loops
The loops in update_one can be increased in two different ways: step
by one for files and by <n> for directories. "for" loop is not
suitable for this as it always steps by one and special handling is
required for directories. Replace them with "while" loops for clarity.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
cache-tree: remove dead i-t-a code in verify_cache()
This code is added in 331fcb5 (git add --intent-to-add: do not let an
empty blob be committed by accident - 2008-11-28) to forbid committing
when i-t-a entries are present. When we allow that, we forgot to
remove this.
Noticed-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
README: it does not matter who the current maintainer is
The audience of this introductory document does not have to know nor
interact with the maintainer, so drop the mention of him. Other
documents such as SubmittingPatches may be a more suitable place to
have it.
In remote-test-svn, there is a parse_rev_note function to
parse lines of the form "Revision-number" from notes. If it
finds such a line and parses it, it returns 0, copying the
value into a "struct rev_note". If it finds an entry that is
garbled or out of range, it returns -1 to signal an error.
However, if it does not find any "Revision-number" line at
all, it returns success but does not put anything into the
rev_note. So upon a successful return, the rev_note may or
may not be initialized, and the caller has no way of
knowing.
gcc does not usually catch the use of the unitialized
variable because the conditional assignment happens in a
separate function from the point of use. However, when
compiling with -O3, gcc will inline parse_rev_note and
notice the problem.
We can fix it by returning "-1" when no note is found (so on
a zero return, we always found a valid value).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Explain that --tags is just like another explicit refspec on the
command line and as such overrides the default refspecs configured
via the remote.$name.fetch variable.
git.txt: add missing info about --git-dir command-line option
Unlike other environment variables (e.g. GIT_WORK_TREE, GIT_NAMESPACE),
the Documentation/git.txt file did not mention that the GIT_DIR
environment variable can also be set using the --git-dir command line
option.
Signed-off-by: Manlio Perillo <manlio.perillo@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
A cache-tree entry with a negative entry count is considered invalid
by the current Git; it records that we do not know the object name
of a tree that would result by writing the directory covered by the
cache-tree as a tree object.
Clarify that any entry with a negative entry count is invalid, but
the implementations must write -1 there. This way, we can later
decide to allow writers to use negative values other than -1 to
encode optional information on such invalidated entries without
harming interoperability; we do not know what will be encoded and
how, so we keep these other negative values as reserved for now.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
1. The original parsed the mailmap file itself, and it did
it wrong (it did not understand entries with an extra
email key).
Instead, this version uses git's "%aE" and "%aN"
formats to have git perform the mapping, meaning we do
not have to read .mailmap at all, but still operate on
the current state that git sees (and it also works
properly from subdirs).
2. The original would find multiple names for an email,
but not the other way around.
This version can do either or both. If we find multiple
emails for a name, the resolution is less obvious than
the other way around. However, it can still be a
starting point for a human to investigate.
3. The original would order only by count, not by recency.
This version can do either. Combined with showing the
counts, it can be easier to decide how to resolve.
4. This version shows similar entries in a blank-delimited
stanza, which makes it more clear which options you are
picking from.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Linus used a lot of different per-machine email addresses in
the early days. This means that "git shortlog -nse" does not
aggregate his counts, and he is listed well below where he
should be (8th instead of 3rd).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit adc3192 (Martin Langhoff has a new e-mail address,
2010-10-05) added a mailmap entry, but forgot that both the
old and new email addresses need to appear for one to be
mapped to the other (i.e., we do not key mailmap emails by
name).
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch updates git's .mailmap in cases where multiple
names are matched to a single email. The "master" name for
each email was chosen by:
1. If the only difference is in the presence or absence
of accented characters, the accented form is chosen
(under the assumption that it is the natural spelling,
and accents are sometimes stripped in email).
2. Otherwise, the most commonly used name is chosen.
3. If all names are equally common, the most recently used name is
chosen.
Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
The description of __git_ps1 function operating in two-arg mode was
not very clear. It said "set PROMPT_COMMAND=__git_ps1" which is not
the right usage for this mode, followed by "To customize the prompt,
do this", giving a false impression that those who do not want to
customize it can get away with no-arg form, which was incorrect.
Make it clear that this mode always takes two arguments, pre and
post, with an example.
The straight-forward one should be listed as the primary usage, and
the confusing one should be an alternate for advanced users. Swap
the order of these two.
Acked-by: Simon Oosthoek <s.oosthoek@xs4all.nl> Signed-off-by: Junio C Hamano <gitster@pobox.com>
For bash completion, the option '-o bashdefault' is used to indicate
that when no other choices are available, file completion should be
performed. Since this option is not available in tcsh, no file
completion is ever performed. Therefore, commands like 'git add ',
'git send-email ', etc, require the user to manually type out
the file name. This can be quite annoying.
To improve the user experience we try to simulate file completion
directly in this script (although not perfectly).
The known issues with the file completion simulation are:
- Possible completions are shown with their directory prefix.
- Completions containing shell variables are not handled.
- Completions with ~ as the first character are not handled.
Signed-off-by: Marc Khouzam <marc.khouzam@ericsson.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
MinGW has a workaround when rmdir unnecessarily fails to retry with
a prompt, but the logic was kicking in when the rmdir failed with
ENOTEMPTY, i.e. was expected to fail and there is no point retrying.
* ef/mingw-rmdir:
mingw_rmdir: do not prompt for retry when non-empty
* ef/mingw-tty-getpass:
mingw: get rid of getpass implementation
mingw: reuse tty-version of git_terminal_prompt
compat/terminal: separate input and output handles
compat/terminal: factor out echo-disabling
mingw: make fgetc raise SIGINT if apropriate
mingw: correct exit-code for SIGALRM's SIG_DFL