Remove the use of the [[ keyword in favour of the POSIX test command
(SC3010).
Don't hard-code pathnames to external programs and define silly
variables such as TMPWATCH. Instead, define a sane PATH at the beginning
of the script.
Don't force the user to reason with whether certain variables need to be
set or not. Instead, call out to portageq in the samples. Further, check
the exit status of portageq, rather than blindly assume that it always
succeeds. As a result, there are no more hard-coded paths, other than
/tmp.
Use pgrep in a less sloppy way.
Query PORT_LOGDIR if the query for PORTAGE_LOGDIR fails.
Closes: https://bugs.gentoo.org/524698
Signed-off-by: Kerin Millar <kfm@plushkava.net>
Signed-off-by: Sam James <sam@gentoo.org>
While we have a dash patch already, it needs some more work
and a bunch of our patches should really be upstreamed anyway.
Need this for a dependency so this works for now.
Signed-off-by: Sam James <sam@gentoo.org>
Unfortunately for the longdescription, it looked quite terrible on
p.g.o as it discards all formatting. Some of this also doesn't reflect
the state of things very well (most clients aren't in the tree and
also dead upstream). For now, opting to simply drop it.
Also reduced amount of apes in the metadata.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
This ebuild has some accumulated history, this is an attempt to
simplify it and remove obsolete baggage. Hopefully no regressions.
Several changes, but for the more notable ones:
* bump to latest xmms2-devel/s4 snapshots and latest waf,
no real changes since 2016 but upstreams a few fixes
* use github directly rather than generate a tarball
* remove upstreamed patches: be-hash + gcc-10 (thanks slyfox)
* remove unnecessary patches: tarball + ffmpeg2
* don't run tests at compile time by using --notests, and also
add a valgrind check to accommodate tinderboxes (bug #807271)
* use makeopts for build, multiprocessing.eclass was there but unused
* replace hard dev-lang/python dep so it doesn't pull wrong slots
* entire DEPEND moved to BDEPEND except cunit
* several other *DEPEND updates after a full review
* use ffmpeg ebuild style static mappings (allows using them in IUSE)
* move "optionals" outside of the server block so client-only features
like perl/python can be used with USE=-server
* drop ruby support, nothing in the tree uses these bindings, using
ruby27/30 triggers severe warnings, and previous ruby-single only
builds bindings for current "any" ruby ignoring RUBY_TARGETS
(if someone needs, say so and will look into proper ruby-ng support)
* add IUSE=tremor and IUSE=gme, available in the tree and works
* make IUSE=alsa non-default, this should be handled by profile/global
* IUSE=mp4 was a no-op and always enabled for good reasons, removed
* install s4 cli tool to manipulate xmms2's database
* enable mid1 with fluidsynth so it can actually load .mid/rmi files
* set default soundfont so it can work out of the box with optfeature
* remove xmms2-et/phonehome, little point with an inactive upstream
(open to re-adding if someone, or the devs, actually want it)
* fix prefix, previously broken due to complete lack of ${EPREFIX}
and also needed an additional fix to find boost includes
Closes: https://bugs.gentoo.org/807271
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
Only tidying patches that will still be using, others
are due for cleanup.
Perl patch was kept in a "need porting" state but
perl-core/ExtUtils-ParseXS is gone from the tree and
so should the issues it was causing.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
Sam James
Ionen Wolkens
Michał Górny
Marek Szuba