| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
| |
If array.fromfile() is not used, then _test_pty_eof() is useless.
This function was for runtime detection of python issue 5380:
http://bugs.python.org/issue5380
However, array.fromfile() use has since been abandoned due to
bugs that exist in all known versions of Python (including Python
2.7 and Python 3.2). See PipeReaderArrayTestCase, for example.
|
|
|
|
|
| |
This helps to avoid triggering the environment.bz2 extraction in
vardbapi.aux_get().
|
| |
|
|
|
|
|
|
| |
We don't want to modify anything outside of the primary DISTDIR,
and symlinks typically point to PORTAGE_RO_DISTDIRS. This will
fix bug #395705.
|
|
|
|
|
| |
This helps to avoid triggering the environment.bz2 extraction in
vardbapi.aux_get().
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
The config class has equivalent code, but we also need to do it here if
_disable_legacy_globals() has been called.
|
| |
|
|
|
|
| |
Files under /var/cache/ are allowed to disappear.
|
| |
|
|
|
|
|
|
|
| |
Since commit a715b65f7bd36409c1283e6911265d1f4405ab7a, the Package.root
and RootConfig.root attributes refer to EROOT instead of ROOT.
Therefore, adjust display code so that EROOT is only displayed when
ROOT != "/".
|
| |
|
| |
|
|
|
|
| |
Also, generate appropriate SKIP message if pty is not acquired.
|
|
|
|
| |
Allows reliable triggering of the failure mode on fast computers.
|
| |
|
| |
|
| |
|
|
|
|
| |
instead of portage.repository.config.RepoConfigLoader.__init__().
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
We have abandoned array.fromfile() due to bugs that exist in all known
versions of Python (including Python 2.7 and Python 3.2). See
PipeReaderArrayTestCase, for example.
|
| |
|
|
|
|
|
| |
Since SpawnProcess no longer uses array, add conditional array support
to PipeReader and use that for tests.
|
| |
|
|
|
|
|
| |
Since commit 30d2d0a9db486c5a70848ad5d27b37a3ec48f271, we use os.read()
due to bugs in array.fromfile(). So, eliminate array usage entirely.
|
|
|
|
|
| |
Since commit 30d2d0a9db486c5a70848ad5d27b37a3ec48f271, we use os.read()
due to bugs in array.fromfile(), so there's no need to use fdopen.
|
|
|
|
|
|
|
|
| |
In PipeReader and AbstractPollTask, we have to treat EIO as EOF, in
order to terminate the select loop at the correct time when using
PollSelectAdapter. This fixes a regression in EIO handling since
commits 915348ce34fc499ac295b8f0ffee9f0829803542 adn
30d2d0a9db486c5a70848ad5d27b37a3ec48f271.
|
|
|
|
| |
https://bugs.pypy.org/issue956
|
|
|
|
|
| |
This makes it unnecessary to explicitly call portage.data._init() in
runTests, and fixes some other cases that trigger the same issue.
|
| |
|
|
|
|
|
|
| |
When I extended test_poll to test SpawnProcess array.fromfile() usage,
it exposed bugs in array.fromfile() that I couldn't find a way to
handle. So, use os.read() instead.
|
| |
|
| |
|
|
|
|
|
|
| |
This allows testing of stage builds with __PORTAGE_TEST_HARDLINK_LOCKS
set in the environment, since __PORTAGE_TEST_HARDLINK_LOCKS needs to
propagate to childs processes for all ROOTs.
|
|
|
|
|
| |
This ensures that global __PORTAGE_TEST_HARDLINK_LOCKS overrides work
correctly for all tests.
|
| |
|
|
|
|
|
|
|
| |
Support for file object or integer file descriptor parameters is
deprecated due to ambiguity in whether or not it's safe to close
the file descriptor, making it prone to "Bad file descriptor" errors
or file descriptor leaks.
|
| |
|
| |
|
|
|
|
|
|
| |
If fstat shows more than one hardlink, then it's extremely unlikely
that the following link call will result in a lock, so optimize away
the wasteful link call and sleep or raise TryAgain.
|
| |
|
|
|
|
|
|
|
|
|
| |
This reverts the behavior change from commit
0cc174b6fc28feb26ea151d76f794e0ff2c2fa39, since Gentoo's council has
voted to revert it in their 20111213 meeting, as requested on the
gentoo-project mailing list:
http://archives.gentoo.org/gentoo-project/msg_4e282bb4e6ac2611de2a39171a803c48.xml
|
|
|
|
|
|
| |
This updates the hardlink locking code to support the non-blocking,
lockfile(wantnewlockfile=False), and lockfile(file_object) behaviors
which are used by portage code.
|
|
|
|
|
| |
This is for consistency with the behavior for normal lock files, since
commit ce44ea3e914098a52bc0d1d995e71661659e77ca (bug #142369).
|
|
|
|
|
|
|
|
| |
Only append the current repo to eclass_locations if it's not there
already. This allows masters to have more control over eclass override
order, which may be useful for scenarios in which there is a plan to
migrate eclasses to a master repo. Thanks to Brian Harring
<ferringb@gentoo.org) for suggesting this behavior.
|