On 12/07/14 13:54, Michał Górny wrote:
Hello, developers.
A quick sit: right now toolchain.eclass is a big blocker for multilib
that doesn't seem to want to fix itself. Considering the complexity of
the eclass, the amount of automagic dependencies and the size of
resulting patches ([1] for a start but it lacks EAPI conditionals), I'm
thinking: wouldn't it be better to just remove awt support completely?
Rationale: gcj doesn't seem to have any future, and has exactly two
reverse dependencies in Gentoo. However, only one of them -- gcj-jdk --
requires awt support. gcj-jdk can be supposedly used as a JDK for old
versions of Java, and supposedly can be used to bootstrap icedtea.
However, it never made it to stable and lags behind gcc. I've opened
a bug requesting lastriting it [2].
If we masked gcj-jdk and USE=awt on gcc, we could successfully continue
working on multilib without having to increase toolchain.eclass
by a few hundred lines. That would probably also be the first step
towards removing gcj, which could become possible once mcpdf is
introduced to replace pdftk [3].
What do you think? I've applied the masks listed here in
non-emul-linux-x86 subprofiles if you'd like to test them (not that
there's anything to test).
[1]:https://511832.bugs.gentoo.org/attachment.cgi?id=389818
[2]:https://bugs.gentoo.org/show_bug.cgi?id=531900
[3]:https://bugs.gentoo.org/show_bug.cgi?id=531898
I agree with this approach.
--
Anthony G. Basile, Ph.D.
Gentoo Linux Developer [Hardened]
E-Mail : bluen...@gentoo.org
GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA
GnuPG ID : F52D4BBA
