otherwise the build fails And please don't tell me to do USE="-vanilla" or set MAKEOPTS="-j1" on the command line. This is an ebuild bug.
Created attachment 356696 [details] openssl-1.0.1c:20130822-165912.log.xz
i'm inclined to not do this. the idea with USE=vanilla is you get the pieces.
Then it must not be stable.
that's illogical
The package does not build.
glibc doesn't build or run for many targets when you use USE=vanilla. that doesn't mean we should remove their KEYWORDS.
(In reply to SpanKY from comment #6) > glibc doesn't build or run for many targets when you use USE=vanilla. > that doesn't mean we should remove their KEYWORDS. I don't see how that invalidates this bug report. Either mask the useflag if it is broken or go fix the bugs. This one is easy to fix. This is a stable package with a valid non-masked useflag and I expect it to work regardless of description or your opinion.
as the meaning of USE=vanilla is clear, it means "you get the pieces" when you disable critical fixes. like build fixes or security changes or ... next you'll complain openssl[vanilla] makes you vulnerable. or doesn't actually build under certain ABIs (regardless of -j level). or breaks with USE=ipv6.
If the useflag is not supported, then it has to be masked.
That's a simple compile failure. Either fix that, or mask the vanilla useflag.
not really. sometimes it works, sometimes it doesn't. if you don't have a good reason for USE=vanilla, then don't set it. that's how it's always been.
(In reply to SpanKY from comment #11) > sometimes it works, sometimes it doesn't. That's your definition of STABLE? > if you don't have a > good reason for USE=vanilla, then don't set it. The reason is some people don't want downstream patches for things like openssl or the kernel (I remember gentoo-sources breaking the e1000e driver). Even build system patches can sometimes break stuff and that happened too (I remember wget). Next time you tell me we shouldn't care about the vanilla-sources ebuild building successfully? The useflag is a good thing and the fix to this bug is a one-liner. > that's how it's always been. Now that argument is definitely illogical.
you could argue that since we make it so easy (and more or less default) for people to use parallel make using the emake wrapper (ebuild and PM), we should also neutralize it's effects for the USE="vanilla" however, i wouldn't go down to using `make` instead of `emake -j1`, since `emake` can contain other useful flags please reconsider, since i'd call adding -j1 same as keeping the upstream behavior
Alternatively, openssl vanilla could be added to profiles/eapi-5-files/package.use.stable.mask.
(In reply to Ulrich Müller from comment #14) > Alternatively, openssl vanilla could be added to > profiles/eapi-5-files/package.use.stable.mask. sounds good (I suggested adding 'vanilla' to eapi-5-files/use.stable.mask earlier at #g-qa, but I have no strong preference if it should be global or not)
*** Bug 491268 has been marked as a duplicate of this bug. ***