* master:
pager: do not fork a pager if PAGER is set to empty.
diff-options: add --patch-with-stat
diff-files --stat: do not dump core with unmerged index.
Support "git cmd --help" syntax
diff --stat: do not do its own three-dashes.
diff-tree: typefix.
GIT v1.3.0-rc4
xdiff: post-process hunks to make them consistent.
This uses the "--no-walk" flag that I never actually implemented (but I'm
sure I mentioned it) to make "git show" be essentially the same thing as
"git whatchanged --no-walk".
It just refuses to add more interesting parents to the revision walking
history, so you don't actually get any history, you just get the commit
you asked for.
I was going to add "--no-walk" as a real argument flag to git-rev-list
too, but I'm not sure anybody actually needs it. Although it might be
useful for porcelain, so I left the door open.
[jc: ported to the unified option structure by Linus]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Merging all three option parsers related to whatchanged is
unarguably the right thing, but the fallout was too big to scare
me away. Let's try it once again, but once step at time.
This splits out init_revisions() call from setup_revisions(), so
that the callers can set different defaults to match the
traditional benaviour.
The rev-list command is still broken in a big way, which is the
topic of next step.
The "--help" argument is special, in that it is (along with "--version")
in that is taken by the "git" program itself rather than the sub-command,
and thus we've had the syntax "git --help cmd".
However, as anybody who has ever used CVS or some similar devil-spawn
program, it's confusing as h*ll when options before the sub-command act
differently from options after the sub-command, so this quick hack just
makes it acceptable to do "git cmd --help" instead, and get the exact same
result.
It may be hacky, but it's simple and does the trick.
Of course, this does not help if you use one of the non-builtin commands
without using the "git" helper. Ie you won't be getting a man-page just
because you do "git-rev-list --help". Don't expect us to be quite _that_
helpful.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
I missed that "git-diff-* --stat" spits out three-dash separator
on its own without being asked. Remove it.
When we output commit log followed by diff, perhaps --patch-with-stat,
for downstream consumer, we _would_ want the three-dash between
the message and the diff material, but that logic belongs to the
caller, not diff generator.
Recent diff_tree_setup_paths() update made it take a second
argument of type "struct diff_options", but we passed another
struct that happenes to have that type at the beginning by
mistake.
I've merged everything I think is ready for 1.3.0, so this is
the final round -- hopefully I can release this with minimum
last-minute fixup as v1.3.0 early next week.
Common option parsing for "git log --diff" and friends
This basically does a few things that are sadly somewhat interdependent,
and nontrivial to split out
- get rid of "struct log_tree_opt"
The fields in "log_tree_opt" are moved into "struct rev_info", and all
users of log_tree_opt are changed to use the rev_info struct instead.
- add the parsing for the log_tree_opt arguments to "setup_revision()"
- make setup_revision set a flag (revs->diff) if the diff-related
arguments were used. This allows "git log" to decide whether it wants
to show diffs or not.
- make setup_revision() also initialize the diffopt part of rev_info
(which we had from before, but we just didn't initialize it)
- make setup_revision() do all the "finishing touches" on it all (it will
do the proper flag combination logic, and call "diff_setup_done()")
Now, that was the easy and straightforward part.
The slightly more involved part is that some of the programs that want to
use the new-and-improved rev_info parsing don't actually want _commits_,
they may want tree'ish arguments instead. That meant that I had to change
setup_revision() to parse the arguments not into the "revs->commits" list,
but into the "revs->pending_objects" list.
Then, when we do "prepare_revision_walk()", we walk that list, and create
the sorted commit list from there.
This actually cleaned some stuff up, but it's the less obvious part of the
patch, and re-organized the "revision.c" logic somewhat. It actually paves
the way for splitting argument parsing _entirely_ out of "revision.c",
since now the argument parsing really is totally independent of the commit
walking: that didn't use to be true, since there was lots of overlap with
get_commit_reference() handling etc, now the _only_ overlap is the shared
(and trivial) "add_pending_object()" thing.
However, I didn't do that file split, just because I wanted the diff
itself to be smaller, and show the actual changes more clearly. If this
gets accepted, I'll do further cleanups then - that includes the file
split, but also using the new infrastructure to do a nicer "git diff" etc.
Even in this form, it actually ends up removing more lines than it adds.
It's nice to note how simple and straightforward this makes the built-in
"git log" command, even though it continues to support all the diff flags
too. It doesn't get much simpler that this.
I think this is worth merging soonish, because it does allow for future
cleanup and even more sharing of code. However, it obviously touches
"revision.c", which is subtle. I've tested that it passes all the tests we
have, and it passes my "looks sane" detector, but somebody else should
also give it a good look-over.
[jc: squashed the original and three "oops this too" updates, with
another fix-up.]
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Clean up trailing whitespace when pretty-printing commits
Partly because we've messed up and now have some commits with trailing
whitespace, but partly because this also just simplifies the code, let's
remove trailing whitespace from the end when pretty-printing commits.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
Relying on eye-candy progress bar was fragile to begin with.
Run fetch-pack with -k option, and count the objects that are in
the pack that were transferred from the other end.
Shell utilities: Guard against expr' magic tokens.
Some words, e.g., `match', are special to expr(1), and cause strange
parsing effects. Track down all uses of expr and mangle the arguments
so that this isn't a problem.
Signed-off-by: Mark Wooding <mdw@distorted.org.uk> Signed-off-by: Junio C Hamano <junkio@cox.net>
t3600-rm: skip failed-remove test when we cannot make an unremovable file.
When running t3600-rm test under fakeroot (or as root), we
cannot make a file unremovable with "chmod a-w .". Detect this
case early and skip that test.
This trivially avoids keeping the commit message data around after we
don't need it any more, avoiding a continually growing "git log" memory
footprint.
It's not a huge deal, but it's somewhat noticeable. For the current kernel
tree, doing a full "git log" I got
ie the touched pages dropped from 8851 to 5039. For the historic kernel
archive, the numbers are 18357->11037 minor page faults.
We could/should in theory free the commits themselves, but that's really a
lot harder, since during revision traversal we may hit the same commit
twice through different children having it as a parent, even after we've
shown it once (when that happens, we'll silently ignore it next time, but
we still need the "struct commit" to know).
And as the commit message data is clearly the biggest part of the commit,
this is the really easy 60% solution.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
This target lists undocumented commands, and/or whose document
is not referenced from the main git documentation.
For now, there are some exceptions I added primarily because I
lack the energy to document them myself:
- merge backends (we should really document them)
- ssh-push/ssh-pull (does anybody still use them?)
- annotate and blame (maybe after one of them eats the other ;-)
* jc/combine:
stripspace: make sure not to leave an incomplete line.
git-commit: do not muck with commit message when no_edit is set.
When showing a commit message, do not lose an incomplete line.
Retire t5501-old-fetch-and-upload test.
combine-diff: type fix.
* master:
stripspace: make sure not to leave an incomplete line.
git-commit: do not muck with commit message when no_edit is set.
When showing a commit message, do not lose an incomplete line.
Retire t5501-old-fetch-and-upload test.
stripspace: make sure not to leave an incomplete line.
When dealing with a commit log message for human consumption, it
never makes sense to keep a log that ends with an incomplete
line, so make it a part of the clean-up process done by
git-stripspace.
Acked-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-commit: do not muck with commit message when no_edit is set.
Spotted by Linus and Darrin Thompson. When we took a commit
message from -F <file> with an incomplete line, we appended "git
status" output, which ended up attaching a lone "#" at the end.
We still need the "do we have anything to commit?" check by
running "status" (which has to know what to do in different
cases with -i/-o/-a), but there is no point appending its output
to the proposed commit message given by the user.
combine-diff: do not lose hunks with only deletion at end.
We used to lose hunks that appear at the end and have only
deletion. This makes sure that the record beyond the end of
file (which holds such deletions) is examined.
Ok this really should be the good version. The option
handling has been reworked to be automation safe.
Currently to import the -mm tree I have to work around
git-apply by using patch. Because some of Andrews
patches in quilt will only apply with fuzz.
I started out implementing a --fuzz option and then I realized
fuzz is not a very safe concept for an automated system. What
you really want is a minimum number of context lines that must
match. This allows policy to be set without knowing how many
lines of context a patch actually provides. By default
the policy remains to match all provided lines of context.
Allowng git-apply to match a restricted set of context makes
it much easier to import the -mm tree into git. I am still only
processing 1.5 to 1.6 patches a second for the 692 patches in
2.6.17-rc1-mm2 is still painful but it does help.
If I just loop through all of Andrews patches in order
and run git-apply --index -C1 I process the entire patchset
in 1m53s or about 6 patches per second. So running
git-mailinfo, git-write-tree, git-commit-tree, and
git-update-ref everytime has a measurable impact,
and shows things can be speeded up even more.
All of these timings were taking on my poor 700Mhz Athlon
with 512MB of ram. So people with fast machiens should
see much better performance.
When a match is found after the number of context are reduced a
warning is generated. Since this is a rare event and possibly
dangerous this seems to make sense. Unless you are patching
a single file the error message is a little bit terse at
the moment, but it should be easy to go back and fix.
I have also updated the documentation for git-apply to reflect
the new -C option that sets the minimum number of context
lines that must match.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
blame and friends: adjust to multiple pathspec change.
This makes things that include revision.h build again.
Blame is also built, but I am not sure how well it works (or how
well it worked to begin with) -- it was relying on tree-diff to
be using whatever pathspec was used the last time, which smells
a bit suspicious.
Without this flag, "git log -p paths..." shows commits that
touch the specified paths, and diffs about the same specified
paths. With this, the full diff is shown for commits that touch
the specified paths.
The way tree-diff was set up assumed we would use only one set
of pathspec during the entire life of the program. Move the
pathspec related static variables out to diff_options structure
so that we can filter commits with one set of paths while show
the actual diffs using different set of paths.
I suspect this breaks blame.c, and makes "git log paths..." to
default to the --full-diff, the latter of which is dealt with
the next commit.
Nobody except diff-stages used it -- the callers instead filtered
the input to diffcore themselves. Make diff-stages do that as
well and retire diffcore-pathspec.
This separates out the part that deals with one-commit diff-tree
(and --stdin form) into a separate log-tree module.
There are two goals with this. The more important one is to be
able to make this part available to "git log --diff", so that we
can have a native "git whatchanged" command. Another is to
simplify the commit log generation part simpler.
The parent rewriting feature caused us to create the whole history in one
go, and then simplify it later, because of how rewrite_parents() had been
written. However, with a little tweaking, it's perfectly possible to do
even that one incrementally.
Right now, this doesn't really much matter, because every user of
"--parents" will probably generally _also_ use "--topo-order", which will
cause the old non-incremental behaviour anyway. However, I'm hopeful that
we could make even the topological sort incremental, or at least
_partially_ so (for example, make it incremental up to the first merge).
In the meantime, this at least moves things in the right direction, and
removes a strange special case.
Signed-off-by: Linus Torvalds <torvalds@osdl.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
xdiff/xdiffi.c: fix warnings about possibly uninitialized variables
Compiling this module gave the following warnings (some double dutch!):
xdiff/xdiffi.c: In functie 'xdl_recs_cmp':
xdiff/xdiffi.c:298: let op: 'spl.i1' may be used uninitialized in this function
xdiff/xdiffi.c:298: let op: 'spl.i2' may be used uninitialized in this function
xdiff/xdiffi.c:219: let op: 'fbest1' may be used uninitialized in this function
xdiff/xdiffi.c:219: let op: 'bbest1' may be used uninitialized in this function
A superficial tracking of their usage, without deeper knowledge about the
algorithm, indeed confirms that there are code paths on which these
variables will be used uninitialized. In practice these code paths might never
be reached, but then these fixes will not change the algorithm. If these
code paths are ever reached we now at least have a predictable outcome. And
should the very small performance impact of these initializations be
noticeable, then they should at least be replaced by comments why certain
code paths will never be reached.
Some extra initializations in this patch now fix the warnings.
When a broken pair is matched up by rename detector to be merged
back, we do not want to say it is "dissimilar" with the
similarity index. The output should just say they were changed,
taking the break score left by the earlier diffcore-break run if
any.
* jc/blame:
blame -S <ancestry-file>
Match ofs/cnt types in diff interface.
blame: use built-in xdiff
combine-diff: move the code to parse hunk-header into common library.
combine-diff: refactor built-in xdiff interface.
combine-diff: use built-in xdiff.
* master:
gitk: Fix incorrect invocation of getmergediffline
[PATCH] gitk: Fix searching for filenames in gitk
count-delta: match get_delta_hdr_size() changes.
check patch_delta bounds more carefully
git-svnimport: Don't assume that copied files haven't changed
Don't assume that a file that SVN claims was copied from somewhere
else is bit-for-bit identical with its parent, since SVN allows
changes to copied files before they are committed.
Without this fix, such copy-modify-commit operations causes the
imported file to lack the "modify" part -- that is, we get subtle data
corruption.
Signed-off-by: Karl Hasselström <kha@treskal.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
Jens Axboe noticed that recent "git push" has become very slow
since we made --thin transfer the default.
Thin pack generation to push a handful revisions that touch
relatively small number of paths out of huge tree was stupid; it
registered _everything_ from the excluded revisions. As a
result, "Counting objects" phase was unnecessarily expensive.
This changes the logic to register the blobs and trees from
excluded revisions only for paths we are actually going to send
to the other end.