This adds a couple of tests to cover the following renaming
merge cases:
- one side renames and the other side does not, with and without
content conflicts.
- both side rename to the same path, with and without content
conflicts.
The test setup also prepares a case in which both side rename to
different destination, but currently the code collapses these
destination paths and removes the original path, which may be
wrong. The outcome of this case is not checked by the tests in
this round.
Fredrik points out there is a useful wrapper runProgram() used
everywhere that we can use to feed input into subprocess. Use
it to catch errors from the subprocess; it is a good cleanup as
well.
merge-recursive: leave unmerged entries in the index.
This does two things.
- When one branch renamed and the other branch did not, the
resulting half-merged file in the working tree used to swap
branches around and showed as if renaming side was "ours".
This was confusing and inconsistent (even though the conflict
markers were marked with branch names, it was not a good
enough excuse). This changes the order of arguments to
mergeFile in such a case to make sure we always see "our"
change between <<< and ===, and "their" change between ===
and >>>.
- When both branches renamed to the same path, and when one
branch renamed and the other branch did not, we attempt
mergeFile. When this automerge conflicted, we used to
collapse the index. Now we use update-index --index-info
to inject higher stage entries to leave the index in unmerged
state for these two cases.
What this still does _not_ do is to inject unmerged state into
the index when the structural changes conflict. I have not
thought things through what to do in each case yet, but the
cases this commit cover are the most common ones, so this would
be a good start.
People seem to be getting test failure from t6021 not becuase
git is faulty but because they forgot to install "merge". Check
this and other trivial pilot errors in the first test.
use safe_pipe_capture() or system() over backticks where
shellquoting may have been necessary.
More changes planned, so I'm not touching the parts I'm
planning on replacing entirely.
Signed-off-by: Eric Wong <normalperson@yhbt.net> Signed-off-by: Martin Langhoff <martin@catalyst.net.nz>
This is to catch an error where tests are run without first
building what are being tested. Relying on prefixing $PATH with
the build directory and expect that the PATH mechanism would
find what we just built would silently run an already installed
binaries from the PATH.
Signed-off-by: Alex Riesen <raa.lkml@gmail.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-prune: never lose objects reachable from our refs.
Explicit <head> arguments to git-prune replaces, instead of
extends, the list of heads used for reachability analysis by
fsck-objects. By giving a subset of heads by mistake, objects
reachable only from other heads can be removed, resulting in a
corrupted repository.
This commit stops replacing the list of heads, and makes the
command line arguments to add to them instead for safety.
This is a companion patch to e23eff8be92a2a2cb66b53deef020063cff285ed
commit. The same logic, the same rationale that a comparison
function that returns an int should not just compute a ptrdiff_t
and return it.
Not replacing but always including our own refs may be more
desirable (and unarguably much safer), but at the same time I
have a suspicion that that might be forbidding a useful usage I
haven't thought of, so...
Documentation/git-format-patch.txt: Add --signoff, --check, and long option-names.
The documentation was lacking descriptions for the --signoff and --check
options to git-format-patch. It was also missing the following long
option-names: --output-directory (-o), --numbered (-n), --keep-subject
(-k), --author (-a), --date (-d), and --mbox (-m).
Signed-off-by: Nikolai Weibull <nikolai@bitwi.se> Signed-off-by: Junio C Hamano <junkio@cox.net>
Add documentation for git-revert and git-cherry-pick.
* Added the -e option to the documentation of git-cherry-pick.
* Added the -e and --no-commit option to git-revert.
* Removed redundant case expression for -n as --no-edit (already taken by
--no-commit).
Signed-off-by: Nikolai Weibull <nikolai@bitwi.se> Signed-off-by: Junio C Hamano <junkio@cox.net>
gitk: Work around Tcl's non-standard names for encodings
This uses a table of encoding names and aliases distilled from
http://www.iana.org/assignments/character-sets plus some heuristics
to convert standard encoding names to ones that Tcl recognizes.
update-index: allow --index-info to add higher stages.
The new merge world order tells the merge strategies to leave
the cache unmerged and store the automerge result in the working
tree if automerge is not clean. This was done for the resolve
strategy and recursive strategy when no rename is involved, but
recording a conflicting merge in the rename case could not
easily be done by the recursive strategy.
This commit adds a new input format, in addition to the exsting
two, to "update-index --index-info".
(1) mode SP sha1 TAB path
The first format is what "git-apply --index-info"
reports, and used to reconstruct a partial tree
that is used for phony merge base tree when falling
back on 3-way merge.
(2) mode SP type SP sha1 TAB path
The second format is to stuff git-ls-tree output
into the index file.
(3) mode SP sha1 SP stage TAB path
This format is to put higher order stages into the
index file and matches git-ls-files --stage output.
To place a higher stage entry to the index, the path should
first be removed by feeding a mode=0 entry for the path, and
then feeding necessary input lines in the (3) format.
The first line of the input feeds 0 as the mode to remove the
path; the SHA1 does not matter as long as it is well formatted.
Then the second and third line feeds stage 1 and stage 2 entries
for that path. After the above, we would end up with this:
On AIX, there is no -n option to the system's echo. Instead,
it needs the '\c' control character. We could replace
echo -n "foo"
with
echo -e "foo\c"
but printf is recommended by most man pages. Tested on AIX
5.3, Solaris 8, and Debian.
[jc: futureproofed two instances that uses variable with '%s'
so later feeding different messages would not break things too
easily; others are emitting literal so whoever changes the
literal ought to notice more easily so they are safe.]
Signed-off-by: E. Jason Riedy <ejr@cs.berkeley.edu> Signed-off-by: Junio C Hamano <junkio@cox.net>
> The code looks wrong. It assumes that pointers are no larger than ints.
> If pointers are larger than ints, the code does not necessarily compute
> a consistent ordering and qsort is allowed to do whatever it wants.
>
> Morten
>
> static int compare_object_pointers(const void *a, const void *b)
> {
> const struct object * const *pa = a;
> const struct object * const *pb = b;
> return *pa - *pb;
> }
Added an AIX clause in the Makefile; that clause likely
will be wrong for any AIX pre-5.2, but I can only test
on 5.3. mailinfo.c was missing the compat header file,
and convert-objects.c needs to define a specific
_XOPEN_SOURCE as well as _XOPEN_SOURCE_EXTENDED.
Signed-off-by: E. Jason Riedy <ejr@cs.berkeley.edu> Signed-off-by: Junio C Hamano <junkio@cox.net>
git-merge-one-file: resurrect leading path creation.
Since we do not use git-update-index followed by
git-checkout-index -u to create the half-merged file on
conflicting case anymore, we need to make sure the leading
directories are created here.
Maybe a better solution would be to allow update-index to add to
higher stage, and checkout-index to extract from such, but that
is a change slightly bigger than I would like to have so close
to 1.0, so this should do for now.
Documentaiton (read-tree): update description of 3-way
The merge-one-file used to leave the working tree intact, but
it has long been changed to leave the merge result there since 2a68a8659f7dc55fd285d235ae2d19e7a8116c30 commit.
It was cumbersome to feed hash-object the file '-t' (you could
have said "./-t", though). Teach it '--' that terminates the
option list, like everybody else. There is no way to extract
usage string from the command either, so teach it "--help" as
well.
Not everybody can rely on /bin/sh to be sane, and we support
SHELL_PATH for that. Use it.
mktemp(1) is not used anywhere else in the core git. Do not
introduce dependency on it.
Not everybody's "which" gives a sane return value. For example,
on Solaris 'which XXX' says "no XXX in /usr/bin /bin ..." and
exits with zero status. The lesson here is to never use 'which'
in your scripts.
Signed-off-by: Junio C Hamano <junkio@twinsun.com> Signed-off-by: Junio C Hamano <junkio@cox.net>
This attempts to clean up the way various compatibility
functions are defined and used.
- A new header file, git-compat-util.h, is introduced. This
looks at various NO_XXX and does necessary function name
replacements, equivalent of -Dstrcasestr=gitstrcasestr in the
Makefile.
- Those function name replacements are removed from the Makefile.
- Common features such as usage(), die(), xmalloc() are moved
from cache.h to git-compat-util.h; cache.h includes
git-compat-util.h itself.
gitk: Some improvements for the code for updating the display
This should be more robust in the case that some does "Update" before
the initial drawing is finished. It also avoids having to reset the
list of children for each commit and reconstruct it.
The existing config.mak should satisfy almost everyone... You
can change the prefix and other vars catch the new setting
anyways. I had forgotten that ?= acts as = (lazy value binding)
and as not := (immediate value binding).
Signed-off-by: E. Jason Riedy <ejr@cs.berkeley.edu> Signed-off-by: Junio C Hamano <junkio@cox.net>
Documentation: shared repository management in tutorial.
The branch policy script I outlined was improved and polished by
Carl and posted on the list twice since then. It is a shame not
to pick it up, so replace the original outline in
howto/update-hook-example.txt with the latest from Carl.
Also talk about setting up git-shell to allow git-push/git-fetch
only SSH access to a shared repository host in the tutorial.
Mention documentation pages that talk about update and
post-update hooks from git-push, because a frequently asked
question is "I want X to happen when I push" and people would
not know to look at git-receive-pack documentation until they
understand that is what runs on the other end.
Again, dumb transport clients are too dumb to make use of the
top objects information to make a choice among multiple packs,
so computing these lines are useless for now. We could
resurrect them if needed later. Also dumb transport clients
presumably can do their own approximation by downloading idx
files to see how relevant each pack is for their fetch.
We tried to compute pack interdependency information in
$GIT_DIR/objects/info/packs, hoping that dumb transports would
make use of it when choosing from multiple choice, but that has
never materialized, so stop computing D lines for now.
sha1_file.c: make sure packs in an alternate odb is named properly.
We somehow ended up registering packs in alternate object
directories as "dir/object//pack/pack-*", which confusd the
update-server-info code very badly. Also we did not attempt to
detect a mistake of listing the object directory itself as one
of the alternates. This does not lead to incorrect behaviour,
but is simply wasteful, so try to do so when we are trivially
able to.
git.c: remove excess output for debugging when command is too long.
When the given command name was too long, we exited with a
message with the number of bytes of the final command name
inside parentheses, without saying what that number is. It was
only meant as a debugging aid while development, so remove it.
Document the --(no-)edit switch of git-revert and git-cherry-pick
This switch was not documented properly. I decided not to mention
the --no-edit switch in the git-cherry-pick documentation since
we always default to no editing.
Signed-off-by: Petr Baudis <pasky@suse.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
show-branch: allow glob pattern to name branches to show.
With this, you can say "git-show-branch topic/* master" to show
all the topic branches you have under .git/refs/heads/topic/ and
your master branch. Another example is "git-show-branch --list
v1.0*" to show all the v1.0 tags. You can disambiguate by
saying "heads/topic/*" to show only topic branches if you have
tags under .git/refs/tags/topic/ as well.
This patch aims to freshen up a bit the git-ls-tree documentation. It hints
that the list of paths are in fact patterns to be matched, explains the new
-t, --name-only and --name-status options, corrects the original autorship
information to refer to yours sincerely, corrects several grammar mistakes,
etc.
Since the documentation still deserves some significant work (at least
proper description of the pattern matching), I also added the stub notice.
Signed-off-by: Petr Baudis <pasky@suse.cz> Signed-off-by: Junio C Hamano <junkio@cox.net>
git.c: two fixes, gitsetenv type and off-by-one error.
gitsetenv as implemented in compat/setenv.c takes two const char*
and int; match that.
Also fix an incorrect attempt in prepend_to_path() to
NUL-terminate the string which stuffed the NUL character at one
past the end of allocation, and was not needed to begin with (we
copy the old_path string including the NUL which terminates it).
If you try to push into an empty repository with no ref arguments to
git push, it doesn't do anything and doesn't say anything. This adds a
warning when send-pack isn't going to push anything, so you don't
assume that it silently did what you wanted.
Signed-off-by: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <junkio@cox.net>
[PATCH] daemon.c and path.enter_repo(): revamp path validation.
The whitelist of git-daemon is checked against return value from
enter_repo(), and enter_repo() used to return the value obtained
from getcwd() to avoid directory aliasing issues as discussed
earier (mid October 2005).
Unfortunately, it did not go well as we hoped.
For example, /pub on a kernel.org public machine is a symlink to
its real mountpoint, and it is understandable that the
administrator does not want to adjust the whitelist every time
/pub needs to point at a different partition for storage
allcation or whatever reasons. Being able to keep using
/pub/scm as the whitelist is a desirable property.
So this version of enter_repo() reports what it used to chdir()
and validate, but does not use getcwd() to canonicalize the
directory name. When it sees a user relative path ~user/path,
it internally resolves it to try chdir() there, but it still
reports ~user/path (possibly after appending .git if allowed to
do so, in which case it would report ~user/path.git).
What this means is that if a whitelist wants to allow a user
relative path, it needs to say "~" (for all users) or list user
home directories like "~alice" "~bob". And no, you cannot say
/home if the advertised way to access user home directories are
~alice,~bob, etc. The whole point of this is to avoid
unnecessary aliasing issues.
Anyway, because of this, daemon needs to do a bit more work to
guard itself. Namely, it needs to make sure that the accessor
does not try to exploit its leading path match rule by inserting
/../ in the middle or hanging /.. at the end. I resurrected the
belts and suspender paranoia code HPA did for this purpose.
This check cannot be done in the enter_repo() unconditionally,
because there are valid callers of enter_repo() that want to
honor /../; authorized users coming over ssh to run send-pack
and fetch-pack should be allowed to do so.
There is no setenv() in Solaris 5.8. The trivial calls to
setenv() were replaced by putenv() in a much earlier patch,
but setenv() was used again in git.c. This patch just adds
a compat/setenv.c.
The rule for building git$(X) also needs to include compat.
objects and compiler flags. Those are now in makefile vars
COMPAT_OBJS and COMPAT_CFLAGS.
Signed-off-by: E. Jason Riedy <ejr@cs.berkeley.edu> Signed-off-by: Junio C Hamano <junkio@cox.net>