| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
Caught on sci-biology/hmmer, it was incorrectly reporting for TMPDIR
declared with `unset`.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
New warning for non-POSIX compliant head or tail without -n.
Closes: https://bugs.gentoo.org/558360
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
- check for virtual package defining DEPEND or BDEPEND
- check for virtual package with a single provider across versions
Closes: https://bugs.gentoo.org/744784
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Caught on app-admin/salt, it was incorrectly reporting for TMPDIR
declared with `local` or `export`. This creates different parse tree
with tree-sitter, so it was flagging this wrongly. Fix it, update tests.
With this, the diff for gentoo repo is only for app-admin/salt.
Reported-by: Patrick McLean <chutzpah@gentoo.org>
Resolves: https://github.com/pkgcore/pkgcheck/issues/490
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
New check for catching passing compressed manpages to doman and newman,
and passing compressed info to doinfo.
Resolves: https://github.com/pkgcore/pkgcheck/issues/477
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For sourceforge, the order of regexes was bad, resulting in it catching
`downloads.sourceforge.net` as project under `sourceforge.net` instead,
which is very wrong. Fix the order, and update the test to check for
this case specially.
On the same note, add support for `sourceforge.io` as a site - cause
looks like it also appears in the wild.
Resolves: https://github.com/pkgcore/pkgcheck/issues/488
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
Scans HOMEPAGE and SRC_URI for uris matching regexes using which it
extracts remote-id. Skips already defined remote-id types. Skips URIS
that end with ".diff" or ".patch". Prefers to take remote-id from newer
package versions, in case URL updated.
Resolves: https://github.com/pkgcore/pkgcheck/issues/475
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
Because of undefined order between requests call by the check (as they
are performed in parallel), the test could fail if there was no match
with the expected requests replies.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
When a URL is collected from multiple different attributes, the results
weren't updating the selected attribute, which resulted in wrong report.
Resolves: https://github.com/pkgcore/pkgcheck/issues/403
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Closes: https://github.com/pkgcore/pkgcheck/issues/467
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
When checking for redundant version, also check that on a profile
doesn't exist where the package is visible, but all later versions
aren't, which means that this package isn't redundant on that profile.
Resolves: https://github.com/pkgcore/pkgcheck/issues/465
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
Classes inheriting python-single-r1 should use python_gen_cond_dep
instead of python_gen_any_dep. This commit fixes the issue. This also
fixes for distutils-r1 with single mode on.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
Phase hooks (`{pre,post}_${phase}`) are used by portage to run user's
hooks defined at `bashrc` file. Defining them in ebuilds might cause
unexpected behavior, so we should warn about it.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Add missing check from repoman, for unmatched unmask of atom in
`package.mask` files. This checks for any unmask of package, which isn't
masked in parent profiles.
Resolves: https://github.com/pkgcore/pkgcheck/issues/369
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
- handle various new cases, and a little less permissive regex
- add cases of usage of `ver_rs` and `ver_cut`
- suggest the user with replacement
Resolves: https://github.com/pkgcore/pkgcheck/issues/450
Resolves: https://github.com/pkgcore/pkgcheck/issues/451
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Signed-off-by: Matt Turner <mattst88@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/445
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Signed-off-by: Matt Turner <mattst88@gentoo.org>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/446
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
The `:` command is used for various setting of variables, while not
using the value. As a result, `: ${VAR:=${D}}` is fine, as it won't use
spaced variable, but `foo ${VAR:=${D}}` will use the unquoted result
of D.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
Using reserved variable names, and not only defining, is prohibited.
Also add new exceptions to the list of reserved names.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
Taken from https://devmanual.gentoo.org/ebuild-writing/variables/index.html#ebuild-defined-variables
Reported-by: Sam James <sam@gentoo.org>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
tree-sitter-bash uses type variable_name for the variable names when
unset is used. Add a special case when a variable_name is detected whose
parent is unset_command.
Resolves: https://github.com/pkgcore/pkgcheck/issues/353
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
Detect pure-DEPEND weak blockers.
Resolves: https://github.com/pkgcore/pkgcheck/issues/399
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
Detect USE flags, which don't affect dependencies, and should be (most
likely) be removed because they provide little utility.
Resolves: https://github.com/pkgcore/pkgcheck/issues/414
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
check for matching deps in each call, and for matching use flags, and
for matching dependency type.
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Adapted from dev-lang/rust. We didn't have coverage
over this result at all, so throw it in.
More coverage is always good, but in particular,
we want to know if we stop immediately
aborting on complicated deps. It could be a regression
(we parse for ages and give up -> hang).
Signed-off-by: Sam James <sam@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/423
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Check for whether BDEPEND="${DISTUTILS_DEPS}" is missing
in PEP517 ebuilds with DISTUTILS_OPTIONAL set.
Implementation notes:
* Ended up not merging this w/ existing PythonCheck because of how
awkward it made testing and the edge cases that ended up being added.
* Wanted to fold this into PythonCheck's check_depend(), but
this doesn't fit super well, given we need to:
1. check for DISTUTILS_USE_PEP517 != no;
2. check for DISTUTILS_OPTIONAL;
3. check for ${DISTUTILS_DEPS} rather than a specific
interpreter (don't want to get into copying eclass logic
for specific packages needed per backend).
For the last point (3), we could actually just look for
dev-python/gpep517, which might be good enough.
Fixes: https://github.com/pkgcore/pkgcheck#388
Signed-off-by: Sam James <sam@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/389
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Fixes: fde30b40ee2d54221257dc80e7669cf770ee2d28
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/416
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Signed-off-by: Sam James <sam@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/415
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Since distfiles using commit hashes are snapshots by definition, do not
require explicit ".gh.tar.gz" suffix for them.
Signed-off-by: Michał Górny <mgorny@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/398
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
This is based on the repoman check EbuildQuote that reports instances of
some variables that should be quoted in certain contexts.
See: https://gitweb.gentoo.org/proj/portage.git/tree/repoman/lib/repoman/modules/linechecks/quotes/quotes.py?h=portage-3.0.30
Closes: https://github.com/pkgcore/pkgcheck/issues/363
Signed-off-by: Thomas Bracht Laumann Jespersen <t@laumann.xyz>
Closes: https://github.com/pkgcore/pkgcheck/pull/379
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Michał Górny <mgorny@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add an auxiliary check for invalid remote-id values. At the moment,
it primarily checks for the correct number of path components and some
obvious mistakes. In the future, the regular expressions can be
improved to catch invalid user/project names as well.
This includes tests for common valid and invalid values for the current
set of remote-ids. In addition to that, PkgMetadataXmlEmptyElement
check is extended to cover "empty <remote-id/>" path in the code.
Unfortunately, covering "unknown type" path is non-trivial as it
requires a remote-id type that is allowed by the XML Schema but not
covered by validation (which normally happens only when Schema is
updated).
This also requires bumping the minimal pkgcore version in order to grab
XML Schema updates (adding "osdn" remote type).
|
|
|
|
|
|
|
|
|
|
| |
Add a new check that verifies that Python packages using GitHub archives
use the `.gh.tar.gz` suffix to make them distinct from the sdist
tarballs on PyPI.
Resolves: https://github.com/pkgcore/pkgcheck/issues/382
Closes: https://github.com/pkgcore/pkgcheck/pull/383
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
Limit network checks to ftp, http and https URLs, in order to avoid
DeadUrl reports that are duplicate to e.g. BadHomepage.
Closes: https://github.com/pkgcore/pkgcheck/issues/372
|
|
|
|
|
|
|
|
|
|
|
| |
Based on updated PMS rules [1] around reserved for package manager
function and variable names. Add the check for eclasses and ebuilds.
[1] https://gitweb.gentoo.org/proj/pms.git/commit/?id=260e21bf65f
Bug: https://bugs.gentoo.org/843779
Bug: https://bugs.gentoo.org/843680
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Resolves: https://github.com/pkgcore/pkgcheck/issues/373
Resolves: https://github.com/pkgcore/pkgcheck/issues/374
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Calls to eend should always have an argument, there's already a QA
notice in portage for this.
Signed-off-by: Thomas Bracht Laumann Jespersen <t@laumann.xyz>
Closes: https://github.com/pkgcore/pkgcheck/pull/365
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Extend MetadataVarCheck with a new method checking the LICENSE variable
for other variables. The only exception is LICENSE itself. The check
accepts occurrences of $LICENSE and ${LICENSE}.
Closes: https://github.com/pkgcore/pkgcheck/issues/366
Signed-off-by: Thomas Bracht Laumann Jespersen <t@laumann.xyz>
Closes: https://github.com/pkgcore/pkgcheck/pull/368
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
Bug: https://bugs.gentoo.org/835396
Signed-off-by: Ulrich Müller <ulm@gentoo.org>
Closes: https://github.com/pkgcore/pkgcheck/pull/360
Signed-off-by: Michał Górny <mgorny@gentoo.org>
|
|
|
|
|
| |
Remove the test testing deprecated profile with no replacement.
This was not a valid error, and pkgcore 0.12.8 no longer emits it.
|
| |
|
|
|
|
|
|
| |
The behavior for handling EXPORT_FUNCTIONS before inherit is not
consistent across package managers. Enforce calling it after inherit
to guarantee consistent behavior.
|
|
|
|
|
|
| |
Add an initial test verifying that IndirectInherits is reported
for the most basic test case. Further tests will follow once @PROVIDES
support lands in pkgcore.
|
| |
|
| |
|