Compiling sun-jdk-1.4.0.08-r1 with older glibc dies with: javaws-waitid.c: In function `waitid': javaws-waitid.c:28: error: `WEXITED' undeclared (first use in this function) javaws-waitid.c:28: error: (Each undeclared identifier is reported only once javaws-waitid.c:28: error: for each function it appears in.) WEXITED is declared in bits/waitflags.h (declared in sys/wait.h) of glibc 2.3.5, but is missing from 2.3.4 versions. I wouldn't normally report bugs that pertain to old versions of packages but seeing old glibc's still exist in the tree I thought you might want to know.
I'm confused. We don't have sun-jdk-1.4.0.08-r1 in our tree, and it most certainly has never been compiled from source code. Can you please provide more information on about which package you are compiling?
Ah, you're referring to 1.4.2.08-r1, of course, which indeed contains a tiny program which is compiled during merge time. I guess we should just add >=sys-libs/glibc-2.3.5 on this, unless somebody sees any reason not to? This package is x86-only, and 2.3.5 is stable on x86.
Ah, 1.4.2.08 of course, my mistake, sorry. I fixed the description summary to reflect this. Yes, that would probably be reasonable resolution for x86, if you ask me? I was going to suggest something similar as well, but I didn't remember if there was anything special to remember wrt. depending on glibc. The same fix will probably be needed to add to 1.4.2.07 and 1.4.2.09 as well.
Created attachment 67435 [details] Proposed fix to sun-jdk-1.4.2.08-r1.ebuild (In reply to comment #2) Emerging an update to glibc actually solved the problem for me. It is reasonable to add the ">=sys-libs/glibc-2.3.5" dependency restriction on the ebuild, given the cryptic error message users would get instead (very hard to find out what went wrong, without prior knowledge of glibc). The goal of ">=sys-libs/glibc-2.3.5" is to shield end users from that kind of bugs and it comes at a very low cost (just upgrading glibc). I definitely support the inclusion of ">=sys-libs/glibc-2.3.5" on the ebuild. I've sent a proposed correction to the file, but the devs might review the file name, it doesn't follow standards.
I feel odd adding a dep on >=sys-libs/glibc-2.3.5 since the thing we compile is too get around a bug in javaws when using the newer glibc but if we don't and you update glibc, javaws will be broken so probably still the best solution? (ie will always work)
added the dep so it will 'always' work
*** Bug 108283 has been marked as a duplicate of this bug. ***
This issue is solved upstream for 1.4.2.10. I just filed bug #115772 for this.