| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
| |
According "SPECIFIC FILE DESCRIPTIONS" in `man portage`, it is valid to
have a parent path like `:path/to/profile` where the repo identifier is
missing. This refers to a path in the current repo.
Signed-off-by: James Le Cuirot <chewi@gentoo.org>
Closes: https://github.com/pkgcore/pkgcore/pull/435
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
This change broke ConfiguredTree, since it was implementing the ABC
function through the `__getattr__` magic. While good intentions, reality
hit us.
Reverts: cc6a8d054363561890bb0552f1439a8ace92b367
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Signed-off-by: Vitaly Zdanevich <zdanevich.vitaly@ya.ru>
Closes: https://github.com/pkgcore/pkgcore/pull/434
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This moves the burden of derivatives having to ensure they return
immutables, instead this is handled by the internals of prototype.tree
since it already partially was.
This change basically makes it easier to write generators for
cat/pkg/ver listing.
For existing implementations returning immutable objects this has
no impact; `t=(1,2);assert t is tuple(t)` for example, and the
same holds for frozenset.
Signed-off-by: Brian Harring <ferringb@gmail.com>
|
|
|
|
|
|
|
| |
This was always an abstract class, just typing/ABC didn't
exist when it was written. Leverage modern protections.
Signed-off-by: Brian Harring <ferringb@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
The implementation wasn't properly aliasing _get_* to
the underlying raw_repo, and it should've been filtering
packages and versions according to the filter. This adds that.
This isn't a huge issue, but it breaks API guarantees
for code tring to efficienctly access this sort of info
for a repository stack.
Signed-off-by: Brian Harring <ferringb@gmail.com>
|
|
|
|
|
|
|
| |
In this case, it's not needed beyond being explicit- the
GetAttrProxy redirects _get_categories and such to .raw_repo.
Signed-off-by: Brian Harring <ferringb@gmail.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Suppressing $HOME is necessary for hygenic tests; to avoid
anything in the dev's environment becoming relied upon for
the test passing.
A slightly more salient point for me; every `git commit` test
was asking for my gpg unlock for git commits; obviously annoying,
also obviously unhygenic (I could have modified default git commit
template for example).
Signed-off-by: Brian Harring <ferringb@gmail.com>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
Better matches the order defined in skel.ebuild
A prerequisite for https://github.com/pkgcore/pkgcheck/pull/645
Signed-off-by: Anthony Ryan <anthonyryan1@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/425
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
If the child repository defines no license groups, the previous code would
exit out before it integrated the parents license groups.
This fixes that, and adds tests for it.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/424
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bug #421 captures this; pkgcore was allowing version components
as the package name if it was suffixed by a revision; the fix
for that incorrectly limited a package name that has a trailing
revision, but *no version syntax* in the name.
This commit just refactors the revision check to also force a check
for a leading version if revision-like is found.
Resolves: https://github.com/pkgcore/pkgcore/issues/421
Signed-off-by: Brian Harring <ferringb@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/422
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
It's a minor redundant protection + optimization, but
it's worth ensuring the only spot considered for a slot prefix
is left of any use restrictions.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
The previous logic was just checking for a trailing version component;
it wasn't checking for a trailing revision.
Relates to issue: #419
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The support for multiple levels of categories never was
used, thus it's time to remove it since it also simplifies
things.
Doing this means the original dict functionality of
repo.categories should be replaced w/ a sequence which
this also does.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/417
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This has been broken for a long while, thus I'm renaming
it to `--environment-re` (from `--environment`) in the
process.
Tests will follow in a later PR (pquery test infra needs some
work).
Signed-off-by: Brian Harring <ferringb@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/415
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
A little bit ugly, but does the job.
Resolves: https://github.com/pkgcore/pkgcore/issues/297
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Closes: https://bugs.gentoo.org/846503
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
| |
https://github.com/pkgcore/pkgcore/commit/b1053683ecb8e17ed69e5797c32b83dd48452f59#r125093286
Follows: 8d1a2095e9e3d5892d5eef39725d8bd2bcf74cbe
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
After further consideration, it seems that the requiring .group extension
will make our life easier, with exact format, and save us from backup
files, readme or similar.
https://github.com/pkgcore/pkgcore/pull/412#discussion_r1307738865
Suggested-by: Michał Górny <mgorny@gentoo.org>
Follows: d00711f2d6cbae14a57088ef78caa3daf72069aa
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
| |
This is a special metadata files that can be used to group packages
for stabilization bugs.
Resolves: https://github.com/pkgcore/pkgcore/issues/411
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
| |
Follows: b1053683ecb8e17ed69e5797c32b83dd48452f59
Resolves: https://github.com/pkgcore/pkgdev/issues/151
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Add support for proxy configuration to the fetch_base operation, through
the environment variables ``http_proxy`` and ``https_proxy``. I see no
security concern from passing those variables to the fetcher, as they
are run with the same permissions as the rest of pkgcore, under user
level.
https://github.com/pkgcore/pkgdev/issues/103
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Resolves: https://github.com/pkgcore/pkgdev/issues/145
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
| |
Resolves: https://github.com/pkgcore/pkgcheck/issues/602
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
This reverts commit ef79d4dcc640658b335fd80cd69bc743d566010a.
As discussed on IRC, temporary fix for live users.
Bug: https://github.com/pkgcore/pkgcheck/issues/602
Signed-off-by: Sam James <sam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If all of a package's ebuild files have been masked bad, that package
gets filtered out of the iterator for the previous loop in this
function, so a followup loop over the masked bad packages can help to
emit the unreported errors.
Resolves: https://github.com/pkgcore/pkgdev/issues/55
Co-authored-by: YiFei Zhu <zhuyifei1999@gmail.com>
Authored-by: Eric Huber <echuber2@illinois.edu>
Closes: https://github.com/pkgcore/pkgcore/pull/406
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
| |
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
instead
See the last commit for particulars, or fa90aff05306fb4935604e64645f2d1d2049233e .
This just completes the migration for what I can find.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Closes: https://github.com/pkgcore/pkgcore/pull/398
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The context of this is fa90aff05306fb4935604e64645f2d1d2049233e; the
original 'central' manager exposed objecst in the same attr namespace as
it's own internals. Whilst no conflicts occured, it was possible, so
the configuration objects were moved to .objects to remove the potential.
A compatibility shim was added (CompatConfigManager), but no warnings
were added at the time due to presumably the module not existing.
Either way, it exists, thus add a warning. There are multiple users
in pkgcore's code (fixed in followup commit). Pkgcheck and pkgdev are
clean, but it may be wise to wait a release or two before removing the
compatibility shim. It's been in place 11 years, not like it's going to
kill us to wait another month...
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The previous commit disabled the logic to generate prepend/val/append- nothing used it-
this commit realigns the client consumers to expect a single value unless they're invoked
arg_type="repr".
Essentially, now if you ask the config layer for a value, you don't have to know if it's
a value that is considered 'incremental'- you just get back a python object for what you asked,
ready for usage.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Many moons gone past the original config system tried to support the idea
of incrementally stacked values (x y -x z -y == z). This commit removes
the fundamental core of that functionality; the incremental logic is now implemented
in domain specific layers, rather than being a capability of the config layer.
Note: this just removes the logic. The signature of [before, mid, after] return
is still in place, and will be removed as the API is normalized. In particular,
pconfig still reports on .prepend/.apend, there just no longer is content there.
That behaviour (and signature) change will be a followon commit.
Signed-off-by: Brian Harring <ferringb@gmail.com>
Signed-off-by: Arthur Zamarin <arthurzam@gentoo.org>
|