dcommit didn't handle errors returned by SVN and coped very
poorly with concurrent commits that appear in SVN repository
while dcommit was running. In both cases it left git repository
in inconsistent state: index (which was reset with `git reset
--mixed' after a successful commit to SVN) no longer matched the
checkouted tree, when the following commit failed or needed to be
rebased. See http://bugs.debian.org/676904 for examples.
This patch fixes the issues by:
- introducing error handler for dcommit. The handler will try
to rebase or reset working tree before returning error to the
end user. dcommit_rebase function was extracted out of cmd_dcommit
to ensure consistency between cmd_dcommit and the error handler.
- calling `git reset --mixed' only once after all patches are
successfully committed to SVN. This ensures index is not touched
for most of the time of dcommit run.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Straight cut & paste. That's the last class.
* Make Git::SVN load it on its own, its the only thing that needs it.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
* Load Git command functions on its own.
* Load Git::SVN modules on its own.
Drive by refactorings...
* Use our() instead of use vars.
* Eliminate the auto loading of Git functions.
Signed-off-by: Eric Wong <normalperson@yhbt.net>
* Load Git command functions itself.
* Can't access the git-svn switch lexical any more, but its only used by
Git::SVN::Log so turn it into a Git::SVN::Log global.
* Load Git::SVN as needed. No need to load it always, its only used twice.
* Moved a state variable to the routine it's used for. (Drive by refactoring)
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Except for adding the 1; at the end, this is a straight copy & paste.
Tests still pass, but its doubtful Git::SVN will compile on its own
without git-svn being loaded. Next commit will fix that.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
This means it should be able to load without git-svn being loaded.
* Load Git.pm on its own and all the needed command functions.
* It needs to grab at a git-svn lexical $_prefix representing the --prefix
option. Provide opt_prefix() for that. This is a refactoring artifact.
The prefix should really be passed into Git::SVN->new.
* Unqualify unnecessarily fully qualified globals like
$Git::SVN::default_repo_id.
* Lexically isolate the class just to make sure nothing is leaking out.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Put them in a new module called Git::SVN::Utils. Yeah, not terribly
original and it will be a dumping ground. But its better than having
them in the main git-svn program. At least they can be documented
and tested.
* fatal() is used by many classes.
* Change the $can_compress lexical into a function.
This should be enough to extract Git::SVN.
Signed-off-by: Michael G. Schwern <schwern@pobox.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
git-svn insists on creating the "master" head (unless it exists) on every
"fetch". It is useful that it gets created initially, when no head exists
- users expect this git convention of having a "master" branch on initial
clone.
However creating it when there already is another head does not provide any
value - the ref is never updated, so it just gets stale after a while. Also,
some users find it annoying that it gets recreated, especially when they would
like the git branch names to follow SVN repository branch names. More
background in http://thread.gmane.org/gmane.comp.version-control.git/115030
Make git-svn skip the "master" creation if HEAD already points at a valid head.
This means "master" does get created on initial "clone" but does not get
recreated once a user deletes it.
Also, make post_fetch_checkout work with any head that is pointed to by HEAD,
not just "master".
Also, use fatal error handling consistent with the rest of the program for
post_fetch_checkout.
Signed-off-by: Marcin Owsiany <marcin@owsiany.pl>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Since v1.7.0-rc2~11 (git-svn: persistent memoization, 2010-01-30),
git-svn has maintained some private per-repository caches in
.git/svn/.caches to avoid refetching and recalculating some
mergeinfo-related information with every "git svn fetch".
These caches use the 'nstore' format from the perl core module
Storable, which can be read and written quickly and was designed for
transfer over the wire (the 'n' stands for 'network'). This format is
endianness-independent and independent of floating-point
representation.
Unfortunately the format is *not* independent of the perl version ---
new perl versions will write files that very old perl cannot read.
Worse, the format is not independent of the size of a perl integer.
So if you toggle perl's use64bitint compile-time option, then using
'git svn fetch' on your old repositories produces errors like this:
Byte order is not compatible at ../../lib/Storable.pm (autosplit
into ../../lib/auto/Storable/_retrieve.al) line 380, at
/usr/share/perl/5.12/Memoize/Storable.pm line 21
That is, upgrading perl to a version that uses use64bitint for the
first time makes git-svn suddenly refuse to fetch in existing
repositories. Removing .git/svn/.caches lets git-svn recover.
It's time to switch to a platform independent serializer backend with
better compatibility guarantees. This patch uses YAML::Any.
Other choices were considered:
- thawing data from Data::Dumper involves "eval". Doing that without
creating a security risk is fussy.
- the JSON API works on scalars in memory and doesn't provide a
standard way to serialize straight to disk.
YAML::Any is reasonably fast and has a pleasant API. In most
backends, LoadFile() reads the entire file into a scalar anyway and
converts it as a second step, but having an interface that allows the
deserialization to happen on the fly without a temporary is still a
comfort.
YAML::Any is not a core perl module, so we take care to use it when
and only when it is available. Installations without that module
should fall back to using Storable with all its quirks, keeping their
cache files in
.git/svn/.caches/*.db
Installations with YAML peacefully coexist by keeping a separate set
of cache files in
.git/svn/.caches/*.yaml.
In most cases, switching between is a one-time thing, so it doesn't
seem worth the complication to migrate existing caches.
The upshot: after this patch, as long as YAML::Any is installed you
can move your git repository between machines with different perl
installations and "git svn fetch" will work fine. If you do not have
YAML::Any, the behavior is unchanged (and in particular does not get
any worse).
Reported-by: Sandro Weiser <sandro.weiser@informatik.tu-chemnitz.de>
Reported-by: Bdale Garbee <bdale@gag.com>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
This slices off another 600 or so lines from the frighteningly long
git-svn.perl script.
The Git::SVN::Ra interface is similar enough to SVN::Ra that it is
probably safe to ignore most of its implementation on first reading.
(Documenting or moving functions that do not fit that pattern is left
as an exercise to the interested reader.)
[ew: rebased and fixed conflict against
commit c26ddce86d
(git-svn: platform auth providers are working only on 1.6.15 or newer)]
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
This makes the git-svn script shorter and less scary for beginners to
read through for the first time. Take the opportunity to explain the
purpose and basic interface of the Git::SVN::Editor class while at it.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Regression fix for people with libsvn between 1.6.12 and 1.6.15, on
which we tried to use the non-working platform auth providers.
* jc/svn-auth-providers-unusable-at-1.6.12:
git-svn: platform auth providers are working only on 1.6.15 or newer
Matthijs Kooijman reports that the cut-off point 082afee (git-svn:
use platform specific auth providers, 2012-04-26) set at 1.6.12 to
use this feature safely was incorrect, and it is 1.6.15 instead:
http://svn.apache.org/repos/asf/subversion/trunk/CHANGES
Version 1.6.15
* improve some swig parameter mapping (r984565, r1035745)
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
This patch removes a chunk of code (the Git::SVN::Fetcher consumer of
libsvn's tree delta protocol) from git-svn.perl and documents its
interface so the hurried reader does not have to read that code right
away.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Using names in the Git:: namespace means these cannot conflict with a
hypothetical binding teaching Subversion to interact with git
repositories.
Currently the packages are private to git-svn.perl so the choice of
name isn't likely to make much difference. This change is mainly
meant as preparation for splitting out the packages in question as
modules on the public search path.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
git-svn.perl is very long (around 6500 lines) and although it is
nicely split into modules, some new readers do not even notice --- it
is too distracting to see all this functionality collected in a single
file.
Splitting it into multiple files would make it easier for people
to read individual modules straight through and to experiment with
components separately.
Let's start with Git::SVN::Prompt. For simplicity, we install this as
a module in the standard search path, just like the existing Git and
Git::I18N modules. In the process, add a manpage explaining its
interface and that it is not likely to be useful for other projects to
avoid confusion.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
When git svn rebase is performed after an unpushed merge, the
rebase operation follows both parents and replays both the user's
local commits and those from the merged branch. This is usually
not the intended behavior.
This patch adds support for the --preserve-merges/-p flag which
allows for a better workflow by re-applying merge commits as merges.
[ew: fixed a minor syntax error]
Signed-off-by: Avishay Lavie <avishay.lavie@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
When using a Perl script on a system where "perl" found on user's $PATH
could be ancient or otherwise broken, we allow builders to specify the
path to a good copy of Perl with $PERL_PATH. The gitweb test forgot to
use that Perl when running its test.
By Jeff King (1) and Zbigniew Jędrzejewski-Szmek (1)
* jk/maint-gitweb-test-use-sane-perl:
Consistently use perl from /usr/bin/ for scripts
t/gitweb-lib: use $PERL_PATH to run gitweb
With double-digit version components in SVN::Core::VERSION,
a naive string comparison is no longer sufficient and
a numeric comparison is required for correct results.
This fixes a regression introduced in
commit 082afee621
("git-svn: use platform specific auth providers")
where SVN version "1.6.6" was incorrectly assumed to
be newer than the required "1.6.12" version.
[mk: fix namespace references]
[ew: commit message]
Tested-by: Matthijs Kooijman <matthijs@stdin.nl>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While the majority of scripts use '#!/usr/bin/perl', some use
'#!/usr/bin/env perl'. In the end there is no difference, because the
Makefile rewrites "#!.*perl" with "#!$PERL_PATH" in scripted
Porcelains before installing. Nevertheless, the second form can be
misleading, because it suggests that perl found first in $PATH will be
used.
Suggested-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Zbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On Linux, this makes authentication using passwords from gnome-keyring
and kwallet work (only the former was tested). On Mac OS X, this allows
using the OS X Keychain.
Signed-off-by: Matthijs Kooijman <matthijs@stdin.nl>
Acked-by: Eric Wong <normalperson@yhbt.net>
In HTTP with keep-alive it's not uncommon for the client to notice that
the server decided to stop maintaining the current connection only when
sending a new request. This naturally results in -EPIPE and possibly
SIGPIPE.
The subversion library itself makes no provision for SIGPIPE. Some
combinations of the underlying libraries do (typically SIG_IGN-ing it),
some don't.
Presumably for that reason all subversion commands set SIGPIPE to
SIG_IGN early in their main()-s.
So should we.
This, together with the previous patch, fixes the notorious "git-svn
died of signal 13" problem (see e.g.
http://thread.gmane.org/gmane.comp.version-control.git/134936).
Signed-off-by: Roman Kagan <rkagan@mail.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
In order to maintain consistency of the database mapping svn revision
numbers to git commit ids, rev_map_set() defers signal processing until
it's finished with an append transaction.[*]
The conventional way to achieve this is through sigprocmask(), which is
available in perl in the standard POSIX module.
This is implemented by this patch. One important consequence of it is
that the signal handlers won't be unconditionally set to SIG_DFL anymore
upon the first invocation of rev_map_set() as they used to. As a
result, the signals ignored by git-svn parent will remain ignored;
otherwise the behavior remains the same.
This patch paves the way to ignoring SIGPIPE throughout git-svn which
will be done in the followup patch.
[*] Deferring signals is not enough to ensure the database consistency:
the program may die on SIGKILL or power loss, run out of disk space,
etc. However that's a separate issue that this patch doesn't address.
Signed-off-by: Roman Kagan <rkagan@mail.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
open_or_add_dir checks to see if the directory already exists or not.
If it already exists and is not a directory, then we fail. However,
open_or_add_dir did not previously account for the possibility that the
path did exist as a file, but is deleted in the current commit.
In order to prevent this legitimate case from failing, open_or_add_dir
needs to know what files are deleted in the current commit.
Unfortunately that information has to be plumbed through a couple of
layers.
Signed-off-by: Steven Walter <stevenrwalter@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
If we delete a file and recreate it as a directory in a single commit,
we have to tell the server about the deletion first or else we'll get
"RA layer request failed: Server sent unexpected return value (405
Method Not Allowed) in response to MKCOL request"
Signed-off-by: Steven Walter <stevenrwalter@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Use numerical form of time zone to replace alphabetic time zone
abbreviation generated by "%Z". "%Z" is not portable and contain
ambiguity for many areas. For example, CST could be "Central
Standard Time" (GMT-0600) and "China Standard Time" (GMT+0800).
Alphabetic time zone abbreviation is meant for human readability,
not for specifying a time zone for machines.
Failed case can be illustrated like this in linux shell:
> echo $TZ
Asia/Taipei
> date +%Z
CST
> env TZ=`date +%Z` date
Mon Dec 19 06:03:04 CST 2011
> date
Mon Dec 19 14:03:04 CST 2011
[ew: fixed bad package reference inside Git::SVN::Log]
Signed-off-by: Wei-Yin Chen (陳威尹) <chen.weiyin@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Change git-svn to use git-rev-list(1) instead of git-log(1) since the
latter is porcelain that'll cause "git svn rebase" to fail completely
if log.abbrevCommit is set to true in the configuration.
Without this patch the code will fail to parse a SHA1, and then just
spew a bunch of "Use of uninitialized value $hash in string eq"
warnings at "if ($c && $c eq $hash) { ..." and never do anything
useful.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Helped-by: Jonathan Nieder <jrnieder@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Change an invocation of git-rev-list(1) to not use --no-color,
git-rev-list(1) will always ignore that option and the --color option,
so there's no need to pass it.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
It's legitimate to update the mergeinfo property without
actually changing any files. This can happen when changes are
backported to a branch, and then that branch is merged back
into mainline. We still want to record the updated mergeinfo
for book-keeping.
Signed-off-by: Steven Walter <stevenrwalter@gmail.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
Fixes a problem matching repository URLs, especially those with a '+' in
the URL, such as svn+ssh:// URLs. Parts of the URL were interpreted as
special characters by the regex matching.
Signed-off-by: Ted Percival <ted.percival@quest.com>
Acked-by: Eric Wong <normalperson@yhbt.net>
* pt/mingw-misc-fixes:
t9901: fix line-ending dependency on windows
mingw: ensure sockets are initialized before calling gethostname
mergetools: use the correct tool for Beyond Compare 3 on Windows
t9300: do not run --cat-blob-fd related tests on MinGW
git-svn: On MSYS, escape and quote SVN_SSH also if set by the user
t9001: do not fail only due to CR/LF issues
t1020: disable the pwd test on MinGW
Implement a new --ignore-refs option which specifies a regex of refs
to ignore while importing svn history.
This is a useful supplement to the --ignore-paths option, as that
option only operates on the contents of branches and tags, not the
branches and tags themselves.
Signed-off-by: Michael Olson <mwolson@gnu.org>
Acked-by: Eric Wong <normalperson@yhbt.net>
Allow the user to check the patch set before it is commited to SVN. It is
then possible to accept/discard one patch, accept all, or quit.
This interactive mode is similar with 'git send email' behaviour. However,
'git svn dcommit' returns as soon as one patch is discarded.
Part of the code was taken from git-send-email.perl (see 'ask' function)
Tests several combinations of potential answers to
'git svn dcommit --interactive'. For each of them, test whether patches
were commited to SVN or not.
Thanks-to Eric Wong <normalperson@yhbt.net> for the initial idea.
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Frédéric Heitzmann <frederic.heitzmann@gmail.com>
While GIT_SSH does not require any escaping / quoting (e.g. for paths
containing spaces), SVN_SSH requires it due to its use in a Perl script.
Previously, SVN_SSH has only been escaped and quoted automatically if it
was unset and thus derived from GIT_SSH. For user convenience, do the
escaping and quoting also for a SVN_SSH set by the user. This way, the
user is able to use the same unescaped and unquoted syntax for GIT_SSH
and SVN_SSH.
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A few scripted Porcelain implementations pretend as if the routine to show
their own help messages are triggered upon "git cmd --help", but a command
line parser of "git" will hijack such a request and shows the manpage for
the cmd subcommand.
Leaving the code to handle such input is simply misleading.
Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Allow git-svn to populate the svn:mergeinfo property automatically in
a narrow range of circumstances. Specifically, when dcommitting a
revision with multiple parents, all but (potentially) the first of
which have been committed to SVN in the same repository as the target
of the dcommit.
In this case, the merge info is the union of that given by each of the
parents, plus all changes introduced to the first parent by the other
parents.
In all other cases where a revision to be committed has multiple
parents, cause "git svn dcommit" to raise an error rather than
completing the commit and potentially losing history information in
the upstream SVN repository.
This behavior is disabled by default, and can be enabled by setting
the svn.pushmergeinfo config option.
[ew: minor style changes and manpage merge fix]
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Bryan Jacobs <bjacobs@woti.com>
"svn dcommit --mergeinfo" replaces the svn:mergeinfo property in an
upstream SVN repository with the given text. The svn:mergeinfo
property may contain commits originating on multiple branches,
separated by newlines.
Cause space characters in the mergeinfo to be replaced by newlines,
allowing a user to create history representing multiple branches being
merged into one.
Update the corresponding documentation and add a test for the new
functionality.
Signed-off-by: Bryan Jacobs <bjacobs@woti.com>
Acked-by: Sam Vilain <sam@vilain.net>
Acked-by: Eric Wong <normalperson@yhbt.net>
The matching step in commit 3235b7053c
did not properly account for users of the "rewriteRoot"
configuration parameter.
ref: <CANWsHyfHtr0EaJtNsDK9UTcmb_AbLg-1jUA-0uWJ-nEeNosb7w@mail.gmail.com>
Suggested-by: H Krishnan <hetchkay@gmail.com>
Signed-off-by: Eric Wong <normalperson@yhbt.net>
Adds a --preserve-empty-dirs flag to the clone operation that will detect
empty directories in the remote Subversion repository and create placeholder
files in the corresponding local Git directories. This allows "empty"
directories to exist in the history of a Git repository.
Also adds the --placeholder-file flag to control the name of any placeholder
files created. Default value is ".gitignore".
Signed-off-by: Ray Chen <rchen@cs.umd.edu>
Acked-by: Eric Wong <normalperson@yhbt.net>
If the bottom of a mergeinfo range is a commit that maps to a git root
commit, then it doesn't have a parent. In such a case, use git commit
range "$top_commit" rather than "$bottom_commit^..$top_commit".
[ew: line-wrap at 80 columns]
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Acked-by: Eric Wong <normalperson@yhbt.net>
Add "--" in the "git rev-list" command line so that if there is a bug
and the revisions cannot be found, the error message is a bit less
cryptic.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Acked-by: Eric Wong <normalperson@yhbt.net>
git svn log --show-commit had no tests and, consequently, no attention
by the author of
b1b4755 (git-log: put space after commit mark, 2011-03-10)
who kept git svn log working only without --show-commit.
Introduce a test and fix it.
Reported-by: Bernt Hansen <bernt@norang.ca>
Signed-off-by: Michael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
