The package manager spec does not guarantee that IDEPEND will be
available in pre/postrm.
If missing, then it means there is nothing to do and eselect-wine
already cleaned the last wine version ahead of wine removal.
This could notably happen if a user emerge -C eselect-wine then
attempts to remove wine which would result in a postrm failure given
the package manager won't reinstall it during removal.
This used to be a real problem when wine used "pre"rm given `die`
would abort removal and you're left with an unremovable package,
but for a long time now wine migrated to "post"rm so this was at
most just a minor annoyance.
We can keep `|| die` for the actual call, it's good to strongly
appeal that it failed despite being in a situation where it should
have worked and, as noted, that won't block actual removal.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
- Drop patch (Lex workaround fixed upstream)
- Change to Meson (upstream preferred build system, automake no
longer supported/maintained)
- Change rdepends (meson options file doesn't include pcsc version
dependency)
- Add perl/lex to bdepend (meson requires lex, perl used in
generating some files outside of twinserial support)
- Drop usb use flag support (upstream dropped support for building
without usb in meson)
- Src uri changed to .tar.xz
- License updated (codebase is LGPL but contains GPL programs built
from source in build process)
Closes: https://bugs.gentoo.org/945492
Signed-off-by: Mason Rocha <turret@turret.cyou>
Closes: https://github.com/gentoo/gentoo/pull/40433
Signed-off-by: Sam James <sam@gentoo.org>
add optional features.
remove rdep prometheus-client as it's an optfeature too.
Signed-off-by: Nicolas PARLANT <nicolas.parlant@parhuet.fr>
Signed-off-by: Sam James <sam@gentoo.org>
Per PMS:
> Ebuilds are allowed to call them in pkg_preinst and pkg_postinst.
> Ebuilds may also call them in pkg_prerm and pkg_postrm but must not
> rely on them being available.
Closes: https://bugs.gentoo.org/949225
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>
Per PMS:
> Ebuilds are allowed to call them in pkg_preinst and pkg_postinst.
> Ebuilds may also call them in pkg_prerm and pkg_postrm but must not
> rely on them being available.
Bug: https://bugs.gentoo.org/949225
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>
They can't be run when cross-compiling unless user has qemu static-usr
enabled and used, which we can't check in ebuild. Therefore, we need to
also disallow docs building when cross-compiling for the time being.
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>
Controls debug extensions such as `zig dump-zir` or
`zig build-exe --debug-compile-errors` and can change default when
building Zig itself from `ReleaseFast` to `ReleaseSafe`.
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>
If binfmt_misc hooks are installed for QEMU, then running Zig tests will
fail because Zig attempts to run executables during tests that may be
from foreign architectures. Normally this would raise an "exec format
error" failure which Zig handles but binfmt_misc hooks turn it into a
"missing interpreter" error instead which is not handled.
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>
When `--libc` is passed to `zig build test`, tests that cross-compile to
other OSes and architectures fail because the internal include paths
automatically configured by Zig are overridden by the libc installation
file.
Upstream issue: https://www.github.com/ziglang/zig/issues/22383
Signed-off-by: Eric Joldasov <bratishkaerik@landless-city.net>
Signed-off-by: Sam James <sam@gentoo.org>