Check the can_diff and can_merge functions before deciding whether
to add the tool to the available/unavailable lists. This makes
"--tool-help" context-sensitive so that "git mergetool --tool-help"
displays merge tools only and "git difftool --tool-help" displays
diff tools only.
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
vimdiff and vimdiff2 differ only by their merge command so remove the
logic in the diff command since it's not actually needed.
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make the intent of "--bool" more obvious by returning a direct True
or False value. Convert a couple non-bool users with obvious bool
intent.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Invoke git rev-list directly, avoiding the shell, in
P4Submit and P4Sync. The overhead of starting extra
processes is significant in cygwin; this speeds things
up on that platform.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The extra quoting and double-% are unneeded, just to work
around the shell. Instead, avoid the shell indirection.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On windows, p4 marks un-edited files as read-only. Not only are
they read-only, but also they cannot be deleted. Remove the
read-only attribute before deleting in both the copy and rename
cases.
This also happens in the RCS cleanup code, where a file is marked
to be deleted, but must first be edited to remove adjust the
keyword lines. Make sure it is editable before patching.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This test does a commit that is a pure mode change, submits
it to p4 but causes the submit to fail. It verifies that
the state in p4 as well as the client directory are both
unmodified after the failed submit.
On cygwin, "chmod +x" does nothing, so use the test_chmod
function to modify the index directly too.
Also on cygwin, the executable bit cannot be seen in the
filesystem, so avoid that part of the test. The checks of
p4 state are still valid, though.
Thanks-to: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are some old versions of p4, compiled for cygwin, that
treat read-only files differently.
Normally, a file that is not open is read-only, meaning that
"test -w" on the file is false. This works on unix, and it works
on windows using the NT version of p4. The cygwin version
of p4, though, changes the permissions, but does not set the
windows read-only attribute, so "test -w" returns false.
Notice this oddity and make the tests work, even on cygiwn.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This character is not valid in windows filenames, even though
it can appear in p4 depot paths. Avoid using it in tests on
windows, both mingw and cygwin.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In all clients, even those created on windows, use unix line
endings. This makes it possible to verify file contents without
doing OS-specific comparisons in all the tests.
Tests in t9802-git-p4-filetype.sh are used to make sure that
the other LineEnd options continue to work.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
P4 stores newlines in the depos as \n. By default, git does this
too, both on unix and windows. Test to make sure that this stays
true.
Both git and p4 have mechanisms to use \r\n in the working
directory. Exercise these.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Files of type utf16 are handled with "p4 print" instead of the
normal "p4 -G print" interface due to how the latter does not
produce correct output. See 55aa571 (git-p4: handle utf16
filetype properly, 2011-09-17) for details.
On windows, though, "p4 print" can not be told which line
endings to use, as there is no underlying client, and always
chooses crlf, even for utf16 files. Convert the \r\n into \n
when importing utf16 files.
The fix for this is complex, in that the problem is a property
of the NT version of p4. There are old versions of p4 that
were compiled directly for cygwin that should not be subjected
to text replacement. The right check here, then, is to look
at the p4 version, not the OS version. Note also that on cygwin,
platform.system() is "CYGWIN_NT-5.1" or similar, not "Windows".
Add a function to memoize the p4 version string and use it to
check for "/NT", indicating the Windows build of p4.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Replacing \r\n with \n on windows was added in c1f9197 (Replace
\r\n with \n when importing from p4 on Windows, 2007-05-24), to
work around an oddity with "p4 print" on windows. Text files
are printed with "\r\r\n" endings, regardless of whether they
were created on unix or windows, and regardless of the client
LineEnd setting.
As of d2c6dd3 (use p4CmdList() to get file contents in Python
dicts. This is more robust., 2007-05-23), git-p4 uses "p4 -G
print", which generates files in a raw format. As the native
line ending format if p4 is \n, there will be no \r\n in the
raw text.
Actually, it is possible to generate a text file so that the
p4 representation includes embedded \r\n, even though this is not
normal on either windows or unix. In that case the code would
have mistakenly stripped them out, but now they will be left
intact.
More information on how p4 deals with line endings is here:
http://kb.perforce.com/article/63
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Native windows binaries do not understand posix-like
path mapping offered by cygwin. Convert paths to native
using "cygpath --windows" before presenting them to p4d.
This is done using the AltRoots mechanism of p4. Both the
posix and windows forms are put in the client specification,
allowing p4 to find its location by native path even though
the environment reports a different PWD.
Shell operations in tests will use the normal form of $cli,
which will look like a posix path in cygwin, while p4 will
use AltRoots to match against the windows form of the working
directory.
This mechanism also handles the symlink issue that was fixed in
23bd0c9 (git p4 test: use real_path to resolve p4 client
symlinks, 2012-06-27). Now that every p4 client view has
an AltRoots with the real_path in it, explicitly calculating
the real_path elsewhere is not necessary.
Thanks-to: Sebastian Schuberth <sschuberth@gmail.com>
Thanks-to: Johannes Sixt <j6t@kdbg.org>
fixup! git p4 test: translate windows paths for cygwin
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This will avoid having to do native path conversion for
windows. Also may be a bit cleaner always to know that p4d
has that working directory, instead of wherever the function
was called from.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use the standard client_view function from lib-git-p4.sh
instead of building one by hand. This requires a bit of
rework, using the current value of $P4CLIENT for the client
name. It also reorganizes the test to isolate changes to
$P4CLIENT and $cli in a subshell.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The printf command re-interprets the format string as
long as there are arguments to consume. Use this to
simplify a for loop in the client_view() library function.
This requires a fix to one of the client_view callers.
An errant \n in the string was converted into a harmless
newline in the input to "p4 client -i", but now shows up
as a literal \n as passed through by "%s". Remove the \n.
Based-on-patch-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Depot paths must start with //. Exit with a better explanation
when a bad depot path is supplied.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Found by "pyflakes" checker tool.
Modules shelve, getopt were unused.
Module os.path is exported by os.
Reformat one-per-line as is PEP008 suggested style.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit fed2369 (git-p4: Search for parent commit on branch creation,
2012-01-25) uses temporary branches to help find the parent of a
new p4 branch. The temp branches are of the form "git-p4-tmp/%d"
for some p4 change number. Mistakenly, this string was made
using os.path.join() instead of just string concatenation. On
windows, this turns into a backslash (\), which is not allowed in
git branch names.
Reported-by: Casey McGinty <casey.mcginty@gmail.com>
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we look up a sha1 object for reading via parse_object() =>
read_sha1_file() => read_object() callpath, we first check
packfiles, and then loose objects. If we still haven't found it, we
re-scan the list of packfiles in `objects/pack`. This final step
ensures that we can co-exist with a simultaneous repack process
which creates a new pack and then prunes the old object.
This extra re-scan usually does not have a performance impact for
two reasons:
1. If an object is missing, then typically the re-scan will find a
new pack, then no more misses will occur. Or if it truly is
missing, then our next step is usually to die().
2. Re-scanning is cheap enough that we do not even notice.
However, these do not always hold. The assumption in (1) is that the
caller is expecting to find the object. This is usually the case,
but the call to `parse_object` in `everything_local` does not follow
this pattern. It is looking to see whether we have objects that the
remote side is advertising, not something we expect to
have. Therefore if we are fetching from a remote which has many refs
pointing to objects we do not have, we may end up re-scanning the
pack directory many times.
Even with this extra re-scanning, the impact is often not noticeable
due to (2); we just readdir() the packs directory and skip any packs
that are already loaded. However, if there are a large number of
packs, even enumerating the directory can be expensive, especially
if we do it repeatedly.
Having this many packs is a good sign the user should run `git gc`,
but it would still be nice to avoid having to scan the directory at
all.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We generally try to run "gc --auto" after any commands that
might introduce a large number of new objects. An obvious
place to do so is after running "fetch", which may introduce
new loose objects or packs (depending on the size of the
fetch).
While an active developer repository will probably
eventually trigger a "gc --auto" on another action (e.g.,
git-rebase), there are two good reasons why it is nicer to
do it at fetch time:
1. Read-only repositories which track an upstream (e.g., a
continuous integration server which fetches and builds,
but never makes new commits) will accrue loose objects
and small packs, but never coalesce them into a more
efficient larger pack.
2. Fetching is often already perceived to be slow to the
user, since they have to wait on the network. It's much
more pleasant to include a potentially slow auto-gc as
part of the already-long network fetch than in the
middle of productive work with git-rebase or similar.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Python 2.4 lacks the following features:
subprocess.check_call
struct.pack_into
Take a cue from 460d1026 and provide an implementation of the
CalledProcessError exception. Then replace the calls to
subproccess.check_call with calls to subprocess.call that check the return
status and raise a CalledProcessError exception if necessary.
The struct.pack_into in t/9802 can be converted into a single struct.pack
call which is available in Python 2.4.
Signed-off-by: Brandon Casey <bcasey@nvidia.com>
Acked-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Python 2.5 and older do not accept None as the first argument to
translate() and complain with:
TypeError: expected a character buffer object
As suggested by Pete Wyckoff, let's just replace the call to translate()
with a regex search which should be more clear and more portable.
This allows git-p4 to be used with Python 2.5.
Signed-off-by: Brandon Casey <bcasey@nvidia.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Usually a commit that makes it to logmsg_reencode will have
been parsed, and the commit->buffer struct member will be
valid. However, some code paths will free commit buffers
after having used them (for example, the log traversal
machinery will do so to keep memory usage down).
Most of the time this is fine; log should only show a commit
once, and then exits. However, there are some code paths
where this does not work. At least two are known:
1. A commit may be shown as part of a regular ref, and
then it may be shown again as part of a submodule diff
(e.g., if a repo contains refs to both the superproject
and subproject).
2. A notes-cache commit may be shown during "log --all",
and then later used to access a textconv cache during a
diff.
Lazily loading in logmsg_reencode does not necessarily catch
all such cases, but it should catch most of them. Users of
the commit buffer tend to be either parsing for structure
(in which they will call parse_commit, and either we will
already have parsed, or we will load commit->buffer lazily
there), or outputting (either to the user, or fetching a
part of the commit message via format_commit_message). In
the latter case, we should always be using logmsg_reencode
anyway (and typically we do so via the pretty-print
machinery).
If there are any cases that this misses, we can fix them up
to use logmsg_reencode (or handle them on a case-by-case
basis if that is inappropriate).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The logmsg_reencode function will return the reencoded
commit buffer, or NULL if reencoding failed or no reencoding
was necessary. Since every caller then ends up checking for NULL
and just using the commit's original buffer, anyway, we can
be a bit more helpful and just return that buffer when we
would have returned NULL.
Since the resulting string may or may not need to be freed,
we introduce a logmsg_free, which checks whether the buffer
came from the commit object or not (callers either
implemented the same check already, or kept two separate
pointers, one to mark the buffer to be used, and one for the
to-be-freed string).
Pushing this logic into logmsg_* simplifies the callers, and
will let future patches lazily load the commit buffer in a
single place.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When git-commit is asked to reuse a commit message via "-c",
we call read_commit_message, which looks up the commit and
hands back either the re-encoded result, or a copy of the
original. We make a copy in the latter case so that the
ownership semantics of the return value are clear (in either
case, it can be freed).
However, since we return a "const char *", and since the
resulting buffer's lifetime is the same as that of the whole
program, we never bother to free it at all.
Let's just drop the copy. That saves us a copy in the common
case. While it does mean we leak in the re-encode case, it
doesn't matter, since we are relying on program exit to free
the memory anyway.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Replace our use of fnmatch(3) with a more feature-rich wildmatch.
A handful patches at the bottom have been moved to nd/wildmatch to
graduate as part of that branch, before this series solidifies.
We may want to mark USE_WILDMATCH as an experimental curiosity a
bit more clearly (i.e. should not be enabled in production
environment, because it will make the behaviour between builds
unpredictable).
* nd/retire-fnmatch:
Makefile: add USE_WILDMATCH to use wildmatch as fnmatch
wildmatch: advance faster in <asterisk> + <literal> patterns
wildmatch: make a special case for "*/" with FNM_PATHNAME
test-wildmatch: add "perf" command to compare wildmatch and fnmatch
wildmatch: support "no FNM_PATHNAME" mode
wildmatch: make dowild() take arbitrary flags
wildmatch: rename constants and update prototype
Describe tools for automation that were invented since this
document was originally written.
* jc/doc-maintainer:
howto/maintain: document "### match next" convention in jch/pu branch
howto/maintain: mark titles for asciidoc
Documentation: update "howto maintain git"
The "--tool-help" option to git-difftool currently displays incorrect
output since it uses the names of the files in
"$GIT_EXEC_PATH/mergetools/" rather than the list of command names in
git-mergetool--lib.
Fix this by simply delegating the "--tool-help" argument to the
show_tool_help function in git-mergetool--lib.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using show_tool_help from git-difftool we will want it to print
"git difftool" not "git mergetool" so use "git ${TOOL_MODE}tool".
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
TOOL_MODE is set at the top of git-mergetool.sh so there is no need to
set it again in show_tool_help. Removing this lets us re-use
show_tool_help in git-difftool.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is the first step in unifying "git difftool --tool-help" and
"git mergetool --tool-help".
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: David Aguilar <davvid@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An earlier conversion from fgets() to strbuf_getline() in the
codepath to read from /etc/mailname to learn the default host-part
of the ident e-mail address forgot that strbuf_getline() stores the
line at the beginning of the buffer just like fgets().
The "username@" the caller has prepared in the strbuf, expecting the
function to append the host-part to it, was lost because of this.
Reported-by: Mihai Rusu <dizzy@google.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that "already exists" errors are given only when a push tries to
update an existing ref in refs/tags/ hierarchy, we can say "the
tag", instead of "the destination reference", and that is far easier
to understand.
Pointed out by Chris Rorvick.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is harmless in Python 2, which sees the parentheses as redundant
grouping, but is required for Python 3. Since this is the only change
required to make this script just run under Python 3 without needing
2to3 it seems worthwhile.
The case of an empty print must be handled specially because in that
case Python 2 will interpret '()' as an empty tuple and print it as
'()'; inserting an empty string fixes this.
Signed-off-by: John Keeping <john@keeping.me.uk>
Acked-by: Sverre Rabbelier <srabbelier@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Python 3 forbids unbuffered I/O in text mode. Change the reading of
stdin in git-remote-testpy so that we read the lines as bytes and then
decode them a line at a time.
This allows us to keep the I/O unbuffered in order to avoid
reintroducing the bug fixed by commit 7fb8e16 (git-remote-testgit: fix
race when spawning fast-import).
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Under Python 3 'hasher.update(...)' must take a byte string and not a
unicode string. Explicitly encode the argument to this method to hex
bytes so that we don't need to worry about failures to encode that might
occur if we chose a textual encoding.
This changes the directory used by git-remote-testpy for its git mirror
of the remote repository, but this tool should not have any serious
users as it is used primarily to test the Python remote helper
framework.
The use of encode() moves the required Python version forward to 2.0.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The changes to allow this script to run with Python 3 are minimal and do
not affect its functionality on the versions of Python 2 that are
already supported (2.4 onwards).
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Using the approach detailed in the Python documentation[1], run 2to3 on
the code as part of the build if building with Python 3.
The code itself requires no changes to convert cleanly.
[1] http://docs.python.org/3.3/howto/pyporting.html#during-installation
Signed-off-by: John Keeping <john@keeping.me.uk>
Acked-by: Sverre Rabbelier <srabbelier@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When different version of python are used to build via distutils, the
behaviour can change. Detect changes in version and pass --force in
this case.
Signed-off-by: John Keeping <john@keeping.me.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When you have random build artifacts in your build directory, left
behind by running "make" while on another branch, the "git help -a"
command run by __git_list_all_commands in the completion script that
is being tested does not have a way to know that they are not part
of the subcommands this build will ship. Such extra subcommands may
come from the user's $PATH. They will interfere with the tests that
expect a certain prefix to uniquely expand to a known completion.
Instrument the completion script and give it a way for us to tell
what (subset of) subcommands we are going to ship.
Also add a test to "git --help <prefix><TAB>" expansion. It needs
to show not just commands but some selected documentation pages.
Based on an idea by Jeff King.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we push to update an existing ref, if:
* the object at the tip of the remote is not a commit; or
* the object we are pushing is not a commit,
it won't be correct to suggest to fetch, integrate and push again,
as the old and new objects will not "merge". We should explain that
the push must be forced when there is a non-committish object is
involved in such a case.
If we do not have the current object at the tip of the remote, we do
not even know that object, when fetched, is something that can be
merged. In such a case, suggesting to pull first just like
non-fast-forward case may not be technically correct, but in
practice, most such failures are seen when you try to push your work
to a branch without knowing that somebody else already pushed to
update the same branch since you forked, so "pull first" would work
as a suggestion most of the time. And if the object at the tip is
not a commit, "pull first" will fail, without making any permanent
damage. As a side effect, it also makes the error message the user
will get during the next "push" attempt easier to understand, now
the user is aware that a non-commit object is involved.
In these cases, the current code already rejects such a push on the
client end, but we used the same error and advice messages as the
ones used when rejecting a non-fast-forward push, i.e. pull from
there and integrate before pushing again.
Introduce new rejection reasons and reword the messages
appropriately.
[jc: with help by Peff on message details]
Signed-off-by: Junio C Hamano <gitster@pobox.com>
