| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
A broken mess for now which will slowly be turned into a report
generating, semi-equivalent for emerge --info.
|
|
|
|
|
| |
This fixes doing updates that works the resolver such as `pmerge -uD` or
similar.
|
|
|
|
|
|
|
| |
The main benefit is that this gets rid of the 'repository_type' hack
used by pmerge by adding some specific domain attrs for ebuild and
binary repos along with a minor RepositoryGroup enhancement to handle
adding repo groups together which allows us to prioritize repos.
|
|
|
|
|
|
| |
Since new repos aren't synced yet, we can't ascertain if they have a
repo name set in the metadata. Using the repos.conf name for them seems
like a fair compromise until they're pulled down.
|
|
|
|
| |
They must be unique and should generally match repo_ids anyway.
|
| |
|
|
|
|
|
| |
That way we can add other files in there or change the pwrapper* file
names if needed without breaking things.
|
|
|
|
| |
Seems a bit redundant and spaces are annoying as well.
|
|
|
|
|
|
|
| |
This allows us to more easily override the bash version used with a
custom, patched version if trying to debug something. Also, if we want
to support other system types like BSDs in the future its useful as well
since its likely to be found at /usr/local/bin there.
|
| |
|
| |
|
|
|
|
|
|
|
| |
This is probably better than threading conditionals throughout various
functions as they're banned in later EAPIs. The more messy case is
generic deprecation, but the code injection function probably can be
cleaned up more to make things more palatable.
|
| |
|
|
|
|
|
|
| |
Do not force uid/gid when installing binaries to PKGCORE_INST_UID/GID.
Those variables are not set anywhere, and no other helper forces
particular ownership of files (doexe included).
|
|
|
|
|
|
| |
The secpass concept was inherited from Portage but the implementation
disagrees with the current one in Portage. It is unclear and not used
anywhere in pkgcore. Removing it seems the cleanest way forward.
|
|
|
|
|
|
| |
This leads to a bit of code duplication but it'll probably be cleaner
and easier to read as time goes on in order to keep the number of
conditional EAPI specific flags and testing for them at a minimum.
|
|
|
|
|
|
| |
Since we don't use __inject_phase_funcs internally in EAPI specific libs
anymore we don't have to track which functions are defined in external
libs in order to remove them.
|
|
|
|
|
|
|
|
|
|
|
| |
Now all EAPI specific phase functions should be defined in their related
libs using the naming format __phase_$func (e.g. __phase_src_install).
This allows later EAPI definitions to override previous ones and
decreases the complexity for injecting default functions for later
EAPIs.
Also, all __phase_common and __phase_default internal function prefixes
were dropped to just __phase.
|
| |
|
| |
|
|
|
|
|
|
| |
Instead of only during sdist creation. This fixes cases where installing
from live ebuilds or installing via pip without creating an sdist first
wouldn't generate the function lists.
|
|
|
|
|
|
| |
Accidentally pushed test code to master.
This reverts commit 2e9bfdca7882a9fe8d0b45d23b18b55575493957.
|
|
|
|
|
|
| |
Instead of only during sdist creation. This fixes cases where installing
from live ebuilds or installing via pip without creating an sdist first
wouldn't generate the function lists.
|
|
|
|
| |
It's now generated on demand at sdist and install time.
|
|
|
|
|
|
|
| |
Also, remove global function list from git. The scripts to generate the
lists will be run on demand when running tools within the git repo;
otherwise, the generated files will be used if they exist in tarballs
and installed instances.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
We could use declare statements but then we'd have to thread "-g"
arguments into all of them to make them affect the global scope from
within the __var_pop function.
|
| |
|
|
|
|
| |
Fixes #17.
|
|
|
|
| |
Makes it easier to copy them around or change function names later.
|
|
|
|
| |
Patch by Michael Haubenwallner for portage via bug #554084.
|
| |
|
| |
|
|
|
|
|
| |
This still should be fixed to be done lazily on access so we're not
creating duplicate, throw away repo config objs per repo.
|
|
|
|
|
|
| |
Previously specifying a new repo section in repos.conf and syncing
against its section name wouldn't work, only using its location would;
now either should work fine.
|
| |
|
|
|
|
|
| |
It was broken due to allowing multiple repos to be specified since regen
now iterates through them.
|
| |
|
|
|
|
|
| |
Instead of reimplementing basically the same thing in a list
comprehension.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This makes it easier to run all the command line tools under different
configurations, e.g. maintaining separate configurations to target
different binpkg repos.
For portage compat, the path to the custom config directory should be
passed, otherwise if the path to a file is used it will try to load it
as a pkgcore config file.
|
|
|
|
|
|
|
|
|
|
| |
The location parameter for config_from_make_conf() is now the path to
the portage directory (defaults to /etc/portage) instead of just the
main config directory (/etc).
Also, support for setting the system profile via the deprecated
/etc/make.profile symlink has been removed, only
/etc/portage/make.profile is checked.
|
|
|
|
|
|
| |
If a path to a file is passed to load_config() we try to load it as a
pkgcore config file otherwise it gets passed to config_from_make_conf()
as the path to the portage config directory.
|
|
|
|
|
|
| |
At some point we'll probably start storing things under
/usr/share/pkgconfig so it would make sense to have easy access to the
main path as well.
|
| |
|
|
|
|
|
| |
Previously we were duplicating almost the entire make.conf dict into the
fetcher object when it really only requires a few fields.
|