My version of docker (1.8.3) have a different formating for 'docker version'
that broke the build script. We make the version matching more generic in to
work with both version.
Previously we were only checking modified files for their resolve state. But a
file might be unresolved yet not in the modified state. Handle all such cases
properly.
So far pullrebase function has always returned None value, no matter
what orig function returned. This behaviour made impossible for
pull to change returned value from mercurial process (it has always
ended with 0 value by default). This patch makes pullrebase returning
with returned value from orig.
Previously, we'd restore the .orig file after the premerge is complete but
before the merge was complete. This would lead to the .orig file potentially
containing merge conflict markers in it, as a leftover from the last merge
attempt.
sortdict internally maintains a list of keys in insertion order. When a
key is replaced via __setitem__, we .remove() from this list. This
involves a linear scan and array adjustment. This is an expensive
operation.
The tags reading code was calling into sortdict.__setitem__ for each tag
in a read .hgtags revision. For repositories with thousands of tags or
thousands of .hgtags revisions, the overhead from list.remove()
noticeable.
This patch creates a new sortdict() so __setitem__ calls don't incur a
list.remove.
This doesn't appear to have any performance impact on my Firefox
repository. But that's only because tags reading doesn't show up in
profiles to begin with. I'm still waiting to hear from a user with over
10,000 tags and hundreds of heads on the impact of this patch.
4bc805f938a0 made 'bmstore.write()' transaction sensitive, to restore
original bookmarks correctly at failure of a transaction.
For example, shelve and unshelve imply steps below:
before 4bc805f938a0:
1. move active bookmark forward at internal rebasing
2. 'bmstore.write()' writes updated ones into .hg/bookmarks
3. rollback transaction to remove internal commits
4. restore updated bookmarks manually
after 4bc805f938a0:
1. move active bookmark forward at internal rebasing
2. 'bmstore.write()' doesn't write updated ones into .hg/bookmarks
(these are written into .hg/bookmarks.pending, if external hook
is spawn)
3. rollback transaction to remove internal commits
4. .hg/bookmarks should be clean, because it isn't changed while
transaction running: see (2) above
But if shelve or unshelve is executed in the repository created with
"shared bookmarks" ("hg share -B"), this doesn't work as expected,
because:
- share extension makes 'bmstore.write()' write updated bookmarks
into .hg/bookmarks of shared source repository regardless of
transaction activity, and
- intentional transaction failure at the end of shelve/unshelve
doesn't restore already updated .hg/bookmarks of shared source
This patch makes share extension wrap 'bmstore._writerepo()' instead
of 'bmstore.write()', because the former is used to actually write
bookmark changes out.
The cut and head utilities on Solaris have weird differences from the GNU
versions. The f helper script does a dump more nicely than those tools,
anyway.
There are make targets for building mercurial packages for various
distributions using docker. One of the preparation steps before building is to
create inside the docker image a user with the same uid/gid as the current user
on the host system, so that the resulting files have appropriate
ownership/permissions.
It's possible to run `make docker-<distro>` as a user with uid or gid that is
already present in a vanilla docker container of that distibution. For example,
issue4657 is about failing to build fedora packages as a user with uid=999 and
gid=999 because these ids are already used in fedora, and groupadd fails.
useradd would fail too, if the flow ever got to it (and there was a user with
such uid already).
A straightforward (maybe too much) way to fix this is to allow non-unique uid
and gid for the new user and group that get created inside the image. I'm not
sure of the implications of this, but marmoute encouraged me to try and send
this patch for stable.
The _filefoldmap is not updated in when files are deleted from dirstate. In the
case where the file with the same but differently cased name is added afterwards
it renders _filefoldmap incorrect. Those steps must occur to for a problem to
reproduce:
- call status (with listunknown=True),
- update working rectory to a commit which does a casefolding change (A -> a)
- call status again (it will show the file "a" as deleted)
Unfortunately I'm unable to write a test for it because I don't know any
core-mercurial command able to reproduce those steps.
The bug was originally spotted when hgwatchman was enabled. It caused the
changeset contents change during hg rebase (one file unrelarted to changeset
was deleted in it after rebase).
The hgwatchman is able to hit it because when hgignore changes the hgwatchmans
overridestatus is calling original status with listunknown=True.
Multiple threads might attempt to check links with the same temporary
name. This would cause one side to get an EEXIST error and wrongly
fail the support check. Here, we simply retry if our temporary name
exists.
The previous changeset is a simpler way of fixing issue4934 without changing the
spirit of the code. We can remove the dual call to 'delayupdate' but we keep the
tests to show that the issue is still fixed.
The 'prechangegroup' interfere with 'delayupdate' logic because it trigger the
one time call of 'changelog._writepending' (see issure4934). There is no reason
not to call that hook before setting up 'delayupdate' so we move the call a bit
earlier to avoid interference.
The try finally is here to ensure we release the just-created transaction.
Therefore we should not do half a dozen operations before actually entry the try
scope.
As the fromlist gives the names of sub-modules, they should be searched in
the parent directory of the package's __init__.py, which is level=1.
I got the following error by rewriting hgweb to use absolute_import, where
the "mercurial" package is referenced as ".." (level=2):
ValueError: Attempted relative import beyond toplevel package
I know little about the import mechanism, but this change seems correct.
Before this patch, the following code did import the os module with no error:
from mercurial import demandimport
demandimport.enable()
from mercurial import os
print os.name
Because parsers.c does not define PY_SSIZE_T_CLEAN, "s#" format requires
(const char*, int), not (const char*, Py_ssize_t).
https://docs.python.org/2/c-api/arg.html
This error had no problem before 7d13be5f72c2, where datalen wasn't used.
But now fm1readmarkers() fails with "overflow in obsstore" on Python 2.6.9
(amd64) because upper bits of datalen seem to be filled with 1, making it
a negative integer.
This problem seems not visible on our Python 2.7 environment because upper
bits happen to be filled with 0.
Changeset e7b51de6e8eb alters the 'HG_PENDING' mechanism to be "always" there.
This change is made under the assumption than we previously did it only when
"writepending() actually wrote something". This assumption was wrong,
'writepending()' informs of pending changes the first time something is written
and for all following calls. We back this change out to restore the former
behavior, which was already correct.
We need to call delayupdate again after writing to the changelog.
Otherwise the prechangegroup hook consumes the delayupdate subscription and
future hooks don't see the pending changes (see issue 4934 for more details).
Adds a test that triggers the prechangegroup hook before the pretxnchangegroup
hook and verifies that the output of pretxnchangegroup doesn't change.
Previously we would only include HG_PENDING in the hook args if the
transaction's writepending() actually wrote something. This is a bad criteria,
since it's possible that a previous call to writepending() wrote stuff and the
hooks want to still see that.
The solution is to always have hooks execute within the scope of the pending
changes by always putting HG_PENDING in the environment.
The SSH peer class accesses wireproto.commands[cmd] as part of encoding
command arguments. Previously, the wire protocol command was defined in
the clonebundles extension. If the client didn't have this extension
enabled (which it likely doesn't since it is meant as a server-side
extension), then clients attempting to clone via ssh:// would get a
crash due to a KeyError accessing wireproto.commands['clonebundles']
when cloning from a server that is advertising clone bundles.
Moving the definition of the wire protocol command to wireproto.py makes
this problem go away.
A side effect of this code move is servers will always respond to
"clonebundles" wire protocol command requests. This should be fine: the
server will return an empty response unless a clone bundles manifest
file is present and clients shouldn't call the command unless the server
is advertising the capability, which only happens if the clonebundles
extension is enabled and the manifest file exists.
As JSON string is known to be a unicode, we should try round-trip conversion
for localstr type. This patch tests localstr type explicitly because
encoding.fromlocal() may raise Abort for undecodable str, which is probably
not what we want. Maybe we can refactor json filter to use encoding module
more later.
Still "{desc|json}" can't round-trip because showdescription() modifies a
localstr object.
This patch replaces old "DEPRECATED" msgid by "(DEPRECATED)" if that .po
file does not have "(DEPRECATED)" but have "... (DEPRECATED)".
It is necessary to hide deprecated options correctly.
Because f94973e0eefb requires the msgstr of "(DEPRECATED)", old *.po files
must be blamed. Using "DEPRECATED" would just hide the error.
For example, "LANG=da_DK.UTF-8 hg help serve" fails to hide deprecated
options right now, but check-translation.py couldn't detect it because
da.po has outdated translation of "DEPRECATED".
Since f94973e0eefb, deprecated commands, options and so on are
detected by "(DEPRECATED)" instead of "DEPRECATED".
"hg.pot" generated from recent source files doesn't contain msgid
"DEPRECATED", and looking the translation of "DEPRECATED" up in
up-to-date *.po files works incorrectly.
But on the other hand, there are still old *.po files, which contain
msgid "DEPRECATED" but not "(DEPRECATED)". Looking the translation of
"(DEPRECATED)" up in such old *.po files also works incorrectly.
This patch resolves this problem by looking translation of both
"DEPRECATED" and "(DEPRECATED)" up.
This should work correctly, because previous patch makes "deprecated"
checker be applied only on translations, of which msgid contains exact
"(DEPRECATED)" string.
'p.msgstr' examination in 'deprecatedsetup()' is needed to ignore
untranslated entries. This also makes 'deprecatedpe.msgstr'
examination in 'deprecated()' meaningless.
Since f94973e0eefb, deprecated commands, options and so on are
detected by "(DEPRECATED)" instead of "DEPRECATED".
Therefore, 'deprecated' checker in i18n/check-translation.py should
check translation, of which msgid contains "(DEPRECATED)" instead of
"DEPRECATED".
At glance, it seems to do so, but it actually doesn't, because Python
regexp treats "()" as grouping of patterns and "(DEPRECATED)" matches
only against "DEPRECATED".
When using 'extdiff --patch' to check the changes in a rebase, 'precursors(x)'
evaluated to an empty set because I forgot the --hidden flag, so the other
revision was used as the replacement for the empty set. The result was the
patch for the other revision was diffed against itself, and the tool saying
there were no differences. That's misleading since the expected diff args were
silently changed, so it's better to bail out.
The other uses of scmutil.revpair() are commands.diff and commands.status, and
it doesn't make sense to allow an empty revision there either. The code here
was suggested by Yuya Nishihara.
Previously the -rc in our rc tags got dropped, meaning that those
packages looked newer to the packaging system than the later release
build. This rectifies the issue, though some damage may already have
been done on 3.6-rc builds.
I'm mostly cargo-culting the RPM version format - there don't appear
to be rules for RPM about how to handle this. Hopefully an RPM
enthusiast can fix up what I've done as a followup.
'repo.invalidate()' deletes 'filecache'-ed properties by
'filecache.__delete__()' below via 'delattr(unfiltered, k)'. But
cached objects are still kept in 'repo._filecache'.
def __delete__(self, obj):
try:
del obj.__dict__[self.name]
except KeyError:
raise AttributeError(self.name)
If 'repo' object is reused even after failure of command execution,
referring 'filecache'-ed property may reuse one kept in
'repo._filecache', even if reloading from a file is expected.
Executing command sequence on command server is a typical case of this
situation (e0a0f9ad3e4c also tried to fix this issue). For example:
1. start a command execution
2. 'changelog.delayupdate()' is invoked in a transaction scope
This replaces own 'opener' by '_divertopener()' for additional
accessing to '00changelog.i.a' (aka "pending file").
3. transaction is aborted, and command (1) execution is ended
After 'repo.invalidate()' at releasing store lock, changelog
object above (= 'opener' of it is still replaced) is deleted from
'repo.__dict__', but still kept in 'repo._filecache'.
4. start next command execution with same 'repo'
5. referring 'repo.changelog' may reuse changelog object kept in
'repo._filecache' according to timestamp of '00changelog.i'
'00changelog.i' is truncated at transaction failure (even though
this truncation is unintentional one, as described later), and
'st_mtime' of it is changed. But 'st_mtime' doesn't have enough
resolution to always detect this truncation, and invalid
changelog object kept in 'repo._filecache' is reused
occasionally.
Then, "No such file or directory" error occurs for
'00changelog.i.a', which is already removed at (3).
This patch discards objects in '_filecache' other than dirstate at
transaction failure.
Changes in 'invalidate()' can't be simplified by 'self._filecache =
{}', because 'invalidate()' should keep dirstate in 'self._filecache'
'repo.invalidate()' at "hg qpush" failure is removed in this patch,
because now it is redundant.
This patch doesn't make 'repo.invalidate()' always discard objects in
'_filecache', because 'repo.invalidate()' is invoked also at unlocking
store lock.
- "always discard objects in filecache at unlocking" may cause
serious performance problem for subsequent procedures at normal
execution
- but it is impossible to "discard objects in filecache at unlocking
only at failure", because 'releasefn' of lock can't know whether a
lock scope is terminated normally or not
BTW, using "with" statement described in PEP343 for lock may
resolve this ?
After this patch, truncation of '00changelog.i' still occurs at
transaction failure, even though newly added revisions exist only in
'00changelog.i.a' and size of '00changelog.i' isn't changed by this
truncation.
Updating 'st_mtime' of '00changelog.i' implied by this redundant
truncation also affects cache behavior as described above.
This will be fixed by dropping '00changelog.i' at aborting from the
list of files to be truncated in transaction.
Before bundle2, hook output from hook failures was prefixed with
"remote: ". Up to this point with bundle2, the output was converted to
the message to print in an Abort exception. This had 2 implications:
1) It was unclear whether an error message came from the local repo
or the remote
2) The exit code changed from 1 to 255
This patch changes the handling of error:abort bundle2 parts during push
to prefix the error message with "remote: ". This restores the old
behavior.
We still preserve the behavior of raising an Abort during bundle2
application failure. This is a regression from pre-bundle2 because the
exit code changed.
Because we no longer raise an Abort with the remote's message, we needed
to insert a message for the new Abort. So, I invented a new error
message for that. This is another change from pre-bundle2. However, I
like the new error message because it states unambiguously who aborted
the push failed, which I think is important for users so they can decide
what's next.
The added tests don't agree in their output. This demonstrates a
difference in `hg push` behavior between pre-bundle2 and bundle2.
A subsequent patch will attempt to restore some of the pre-bundle2
behavior to bundle2.
I have no idea where it came from, but my clone of Mercurial has an
empty filelog for `contrib/hgfixes/__init__.py` - it's *valid*, just
contains no nodes. Without this change, debugrevlog crashes with a
zero division error.
'unexpected putlfile response: None' when an http error occurs is not very
helpful.
Instead, leave the handling of urllib2.HTTPError exceptions to other layers.
If the store somehow got corrupted, users could end up in weird situations that
were very hard to recover from or lead to propagation of the corruption.
Instead, spend the extra time checking the hash when copying to the working
directory. If it doesn't match, emit a warning, and don't put wrong content in
the working directory.
