The semantics used when parsing notes trees (with regards to fanout subtrees)
follow Dscho's proposal fairly closely:
- No concatenation/merging of notes is performed. If there are several notes
objects referencing a given commit, only one of those objects are used.
- If a notes object for a given commit is present in the "root" notes tree,
no subtrees are consulted; the object in the root tree is used directly.
- If there are more than one subtree that prefix-matches the given commit,
only the subtree with the longest matching prefix is consulted. This
means that if the given commit is e.g. "deadbeef", and the notes tree have
subtrees "de" and "dead", then the following paths in the notes tree are
searched: "deadbeef", "dead/beef". Note that "de/adbeef" is NOT searched.
- Fanout directories (subtrees) must references a whole number of bytes
from the SHA1 sum they subdivide. E.g. subtrees "dead" and "de" are
acceptable; "d" and "dea" are not.
- Multiple levels of fanout are allowed. All the above rules apply
recursively. E.g. "de/adbeef" is preferred over "de/adbe/ef", etc.
This patch changes the in-memory datastructure for holding parsed notes:
Instead of holding all note (and subtree) entries in a hash table, a
simple 16-tree structure is used instead. The tree structure consists of
16-arrays as internal nodes, and note/subtree entries as leaf nodes. The
tree is traversed by indexing subsequent nibbles of the search key until
a leaf node is encountered. If a subtree entry is encountered while
searching for a note, the subtree is unpacked into the 16-tree structure,
and the search continues into that subtree.
The new algorithm performs significantly better in the cases where only
a fraction of the notes need to be looked up (this is assumed to be the
common case for notes lookup). The new code even performs marginally
better in the worst case (where _all_ the notes are looked up).
In addition to this, comes the massive performance win associated with
organizing the notes tree according to some fanout scheme. Even a simple
2/38 fanout scheme is dramatically quicker to traverse (going from tens of
seconds to sub-second runtimes).
As for memory usage, the new code is marginally better than the old code in
the worst case, but in the case of looking up only some notes from a notes
tree with proper fanout, the new code uses only a small fraction of the
memory needed to hold the entire notes tree.
However, there is one casualty of this patch. The old notes lookup code was
able to parse notes that were associated with non-SHA1s (e.g. refs). The new
code requires the referenced object to be named by a SHA1 sum. Still, this
is not considered a major setback, since the notes infrastructure was not
originally intended to annotate objects outside the Git object database.
Cc: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There's no need to be rude to memory-concious callers...
This patch has been improved by the following contributions:
- Junio C Hamano: avoid old-style declaration
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This patch adds the following flags to get_commit_notes() for adjusting the
format of the produced note string:
- NOTES_SHOW_HEADER: Print "Notes:" line before the notes contents
- NOTES_INDENT: Indent notes contents by 4 spaces
Suggested-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Creating repos with 10/100/1000/10000 commits and notes takes a lot of time.
However, using git-fast-import to do the job is a lot more efficient than
using plumbing commands to do the same.
This patch decreases the overall run-time of this test on my machine from
~3 to ~1 minutes.
Signed-off-by: Johan Herland <johan@herland.net>
Acked-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce a 'notemodify' subcommand of the 'commit' command. This subcommand
is similar to 'filemodify', except that no mode is supplied (all notes have
mode 0644), and the path is set to the hex SHA1 of the given "comittish".
This enables fast import of note objects along with their associated commits,
since the notes can now be named using the mark references of their
corresponding commits.
The patch also includes a test case of the added functionality.
Signed-off-by: Johan Herland <johan@herland.net>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "-m" and "-F" options are already the established method
(in both git-commit and git-tag) to specify a commit/tag message
without invoking the editor. This patch teaches "git notes edit"
to respect the same options for specifying a notes message without
invoking the editor.
Multiple "-m" and/or "-F" options are concatenated as separate
paragraphs.
The patch also updates the "git notes" documentation and adds
selftests for the new functionality. Unfortunately, the added
selftests include a couple of lines with trailing whitespace
(without these the test will fail). This may cause git to warn
about "whitespace errors".
This patch has been improved by the following contributions:
- Thomas Rast: fix trailing whitespace in t3301
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git-notes have the potential of being pretty expensive, so test with
a lot of commits. A lot. So to make things cheaper, you have to
opt-in explicitely, by setting the environment variable
GIT_NOTES_TIMING_TESTS.
This patch has been improved by the following contributions:
- Junio C Hamano: tests: fix "export var=val"
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To avoid looking up each and every commit in the notes ref's tree
object, which is very expensive, speed things up by slurping the tree
object's contents into a hash_map.
The idea for the hashmap singleton is from David Reiss, initial
benchmarking by Jeff King.
Note: the implementation allows for arbitrary entries in the notes
tree object, ignoring those that do not reference a valid object. This
allows you to annotate arbitrary branches, or objects.
This patch has been improved by the following contributions:
- Junio C Hamano: fixed an obvious error in initialize_hash_map()
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The script 'git notes' allows you to edit and show commit notes, by
calling either
git notes show <commit>
or
git notes edit <commit>
This patch has been improved by the following contributions:
- Tor Arne Vestbø: fix printing of multi-line notes
- Michael J Gruber: test and handle empty notes gracefully
- Thomas Rast:
- only clean up message file when editing
- use GIT_EDITOR and core.editor over VISUAL/EDITOR
- t3301: fix confusing quoting in test for valid notes ref
- t3301: use test_must_fail instead of !
- refuse to edit notes outside refs/notes/
- Junio C Hamano: tests: fix "export var=val"
- Christian Couder: documentation: fix 'linkgit' macro in "git-notes.txt"
- Johan Herland: minor cleanup and bugfixing in git-notes.sh (v2)
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Tor Arne Vestbø <tavestbo@trolltech.com>
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit notes are blobs which are shown together with the commit
message. These blobs are taken from the notes ref, which you can
configure by the config variable core.notesRef, which in turn can
be overridden by the environment variable GIT_NOTES_REF.
The notes ref is a branch which contains "files" whose names are
the names of the corresponding commits (i.e. the SHA-1).
The rationale for putting this information into a ref is this: we
want to be able to fetch and possibly union-merge the notes,
maybe even look at the date when a note was introduced, and we
want to store them efficiently together with the other objects.
This patch has been improved by the following contributions:
- Thomas Rast: fix core.notesRef documentation
- Tor Arne Vestbø: fix printing of multi-line notes
- Alex Riesen: Using char array instead of char pointer costs less BSS
- Johan Herland: Plug leak when msg is good, but msglen or type causes return
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Tor Arne Vestbø <tavestbo@trolltech.com>
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
get_commit_notes(): Plug memory leak when 'if' triggers, but not because of read_sha1_file() failure
When saying the initial branch is equal to the currently active
remote branch, it is probably intended that the branch heads
point to the same commit. Maybe it would be more useful to a
new user to emphasize that the tree contents and history are the
same.
More important, probably, is that this new branch is set up so
that "git pull" merges changes from the corresponding remote
branch. The next paragraph addresses that directly. What the
reader needs to know to begin with is that (1) the initial branch
is your own; if you do not pull, it won't get updated, and that
(2) the initial branch starts out at the same commit as the
upstream.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Björn Gustavsson <bgustavsson@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Aliases with newlines have been a problem since commit 56fc25f (Bash
completion support for remotes in .git/config., 2006-11-05). The chance
of the problem occurring has been slim at best, until commit 518ef8f
(completion: Replace config --list with --get-regexp, 2009-09-11)
removed the case statement introduced by commit 56fc25f. Before removing
the case statement, most aliases with newlines would work unless they
were specially crafted as follows
[alias]
foo = "log -1 --pretty='format:%s\nalias.error=broken'"
After removing the case statement, a more benign alias like
[alias]
whowhat = "log -1 --pretty='format:%an <%ae>\n%s'"
wont-complete = ...
would cause the completion to break badly.
For now, revert the removal of the case statement until someone comes up
with a better way to get keys from git-config.
Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After commit 511a3fc (wrap git's main usage string., 2009-09-12), the
bash completion for git commands includes COMMAND and [ARGS] when it
shouldn't. Fix this by grepping more strictly for a line with git
commands. It's doubtful whether git will ever have commands starting
with anything besides numbers and letters so this should be fine. At
least by being stricter we'll know when we break the completion earlier.
Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Acked-by: Shawn O. Pearce <spearce@spearce.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This fixes an obvious syntax error that snuck in commit 7e787953:
syntax error at /home/ingmar/bin//git-import-tars line 143, near "/^$/ { "
syntax error at /home/ingmar/bin//git-import-tars line 145, near "} else"
syntax error at /home/ingmar/bin//git-import-tars line 152, near "}"
Signed-off-by: Ingmar Vanhassel <ingmar@exherbo.org>
Acked-and-Tested-by: Peter Krefting <peter@softwolves.pp.se>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Idealists may want USE_NSEC to be the default on Linux some day.
Point to a patch to better explain the requirements on
filesystem code for that to happen.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is not necessarily obvious to a git novice what it means for a
filesystem tree to be equal to the HEAD. Spell it out.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The documentation seems to assume that the starting point for a new
branch is the tip of an existing (ordinary) branch, but that is not
the most common case. More often, "git branch" is used to begin
a branch from a remote-tracking branch, a tag, or an interesting
commit (e.g. origin/pu^2). Clarify the language so it can apply
to these cases. Thanks to Sean Estabrooks for the wording.
Also add a pointer to the user's manual for the bewildered.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update the documentation for --merged and --no-merged to explain
the meaning of the optional parameter introduced in commit 049716b
(branch --merged/--no-merged: allow specifying arbitrary commit,
2008-07-08).
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Sounds better this way, at least to my ears. ("The syntax and
supported options of git merge" is a plural noun. "the same"
instead of "equal" sounds less technical and seems to convey
the meaning better here.)
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The fmt-merge-message builtin can be invoked as "git fmt-merge-msg" rather
than through the hard link in GIT_EXEC_PATH. Although this is unlikely to
confuse most script writers, it should not hurt to make the documentation
a little clearer anyway.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There is excellent documentation for these options in
Documentation/Makefile, but some users may never find it.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are several reasons a git-pull invocation might not
have anything marked for merge:
1. We're not on a branch, so there is no branch
configuration.
2. We're on a branch, but there is no configuration for
this branch.
3. We fetched from the configured remote, but the
configured branch to merge didn't get fetched (either
it doesn't exist, or wasn't part of the fetch refspec).
4. We fetched from the non-default remote, but didn't
specify a branch to merge. We can't use the configured
one because it applies to the default remote.
5. We fetched from a specified remote, and a refspec was
given, but it ended up not fetching anything (this is
actually hard to do; if the refspec points to a remote
branch and it doesn't exist, then fetch will fail and
we never make it to this code path. But if you provide
a wildcard refspec like
refs/bogus/*:refs/remotes/origin/*
then you can see this failure).
We have handled (1) and (2) for some time. Recently, commit
a6dbf88 added code to handle case (3).
This patch handles cases (4) and (5), which previously just
fell under other cases, producing a confusing message.
While we're at it, let's rewrap the text for case (3), which
looks terribly ugly as it is.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
After doing a rebase, git-svn checks that the SVN URL
is what it expects. However, it does not account for
rewriteRoot, which is a legitimate way for the URL
to change. This produces a lot of spurious errors.
[ew: fixed line wrapping]
Signed-off-by: Alexander Gavrilov <angavrilov@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
When git is compiled with the MIPSpro 7.4.4m compiler, and NO_PTHREADS is
set, and NO_MMAP is _not_ set, then git segfaults when trying to access the
first entry in a reflog. If NO_PTHREADS is not set (which implies that the
pthread library is linked in), or NO_MMAP _is_ set, then the segfault is
not encountered. The conservative choice has been made to set NO_MMAP in
the Makefile to avoid this flaw. The GNU C compiler does not produce this
behavior.
The segfault happens in refs.c:read_ref_at(). The mmap succeeds, and the
loop is executed properly until rec is rewound into the first line (reflog
entry) of the file. The segfault is caught by test 28 of
t1400-update-ref.sh which fails when 'git rev-parse --verify "master@{May 25
2005}"' is called.
So, add a comment in the Makefile to describe why NO_MMAP is set and as a
hint to those who may be interested in unsetting it.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When ls-files was called with -i but no exclude pattern, it was
calling fprintf(stderr, "...", NULL) and then exiting. On Solaris,
passing NULL into fprintf was causing a segfault. On glibc systems,
it was simply producing incorrect output (eg: "(null)": ...). The
NULL pointer was a result of argv[0] not being preserved by the option
parser. Instead of requesting that the option parser preserve
argv[0], use die() with a constant string.
A trigger for this bug was: `git ls-files -i`
Signed-off-by: Ben Walton <bwalton@artsci.utoronto.ca>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since commit dcda3614 removed the use of a variable length array from
builtin-pack-objects.c, it is now safe to compile with the threaded delta
search feature enabled. Formerly, the MIPSpro 7.4.4m compiler warned that
variable length arrays should not be used with pthreads.
Signed-off-by: Brandon Casey <casey@nrlssc.navy.mil>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
During an MSVC build on cygwin, the make program did not notice
when the compiler or linker exited with an error. This was caused
by the scripts exiting with the value returned by system() directly.
On POSIX-like systems, such as cygwin, the return value of system()
has the exit code of the executed command encoded in the first byte
(ie the value is shifted up by 8 bits). This allows the bottom
7 bits to contain the signal number of a terminated process, while
the eighth bit indicates whether a core-dump was produced. (A value
of -1 indicates that the command failed to execute.)
The make program, however, expects the exit code to be encoded in the
bottom byte. Futhermore, it apparently masks off and ignores anything
in the upper bytes.
However, these scripts are (naturally) intended to be used on the
windows platform, where we can not assume POSIX-like semantics from
a perl implementation (eg ActiveState). So, in general, we can not
assume that shifting the return value right by eight will get us
the exit code.
In order to improve portability, we assume that a zero return from
system() indicates success, whereas anything else indicates failure.
Since we don't need to know the exact exit code from the compiler
or linker, we simply exit with 0 (success) or 1 (failure).
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the MSVC section of the Makefile, BASIC_CFLAGS is set to a
value which contains the string "-DWIN32-D_CONSOLE". This results
in a (single) malformed -Define being passed to the compiler.
At least on my cygwin installation, the msvc compiler seems to
ignore this parameter, without issuing an error or warning, and
results in the WIN32 and _CONSOLE macros being undefined. This
breaks the build.
In order to fix the build, we simply insert a space between the
two -Define parameters, "-DWIN32" and "-D_CONSOLE", as originally
intended.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
With --prefix=string that does not end with a slash, the top-level entries
are written out with the specified prefix as expected, but no paths in the
directories are added.
Fix this by adding the prefix in write_archive_entry() instead of letting
get_pathspec() and read_tree_recursive() pair; they are designed to only
handle prefixes that are path components.
Signed-off-by: Rene Scharfe <rene.scharfe@lsrfire.ath.cx>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
git am learned --scissors, git commit learned --dry-run and git log
learned --decorate=long|short recently.
Signed-off-by: Stephen Boyd <bebarino@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When reading the "raw format" timestamp from the input stream, make sure
that the timezone offset is a reasonable value by imitating 7122f82
(date.c: improve guess between timezone offset and year., 2006-06-08).
We _might_ want to also check if the timestamp itself is reasonable, but
that is left for a separate commit.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Back when a74b170 (git-pull: disallow implicit merging to detached HEAD,
2007-01-15) added this check, $? referred to the error status of reading
HEAD as a symbolic-ref; but cd67e4d (Teach 'git pull' about --rebase,
2007-11-28) moved the command away from where the check is, and nobody
noticed the breakage. Ever since, $? has always been 0 (tr at the end of
the pipe to find merge_head never fails) and other case arms were never
reached.
These days, error_on_no_merge_candidates function is prepared to handle a
detached HEAD case, which was what the code this patch removes used to
handle.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Johan Herland
Johannes Schindelin
Junio C Hamano
Simon Arlott
Jonathan Nieder
Björn Gustavsson
Stephen Boyd
Ingmar Vanhassel
Jeff King
Alexander Gavrilov
Brandon Casey
Ben Walton
Ramsay Jones
René Scharfe
Stefan Naewe