ftp://ftp.gnu.org/pub/gnu/glibc/ Looks to me like the only alpha patches we need to keep around are 6020_all_alpha-fix-gcc-4.1-warnings.patch (maybe?) 6028_all_alpha-fix-memchr.patch 6029_all_alpha-fix-memchr.patch 6031_all_alpha-glibc-2.12-epoll_create1.patch 6032_all_alpha-syscall-6.patch
you sure we dont need: 6018_all_alpha-glibc-2.8-cache-shape.patch 6022_alpha_alpha-add-fdatasync-support.patch 6026_all_alpha-fix-rtld-fPIC.patch mips team: ive dropped 6610_all_glibc-r10k-workarounds.patch as it no longer applies. they've rewritten the asm a bit.
(In reply to comment #1) > you sure we dont need: > 6018_all_alpha-glibc-2.8-cache-shape.patch I don't really have a clue what this does, so I guess we should keep it. It definitely didn't go into glibc git. > 6022_alpha_alpha-add-fdatasync-support.patch I'm thinking this isn't needed unless we want to support glibc-2.12 on kernels older than 2.6.22 (when fdatasync was added to alpha). > 6026_all_alpha-fix-rtld-fPIC.patch Aurelien Jarno sent this for review/submission, but it wasn't ever committed. I think we should keep it and try to move it upstream (or figure out a more appropriate fix). http://sourceware.org/ml/libc-ports/2009-07/msg00016.html > mips team: ive dropped 6610_all_glibc-r10k-workarounds.patch as it no longer > applies. they've rewritten the asm a bit. Speaking of mips, the commit message for 6600_mips_librt-mips.patch says """ we kind of screwed ourselves into a corner by having the clock symbols exported only with the glibc-2.0 version ... this patch fixes the export so all new binaries built against librt will use the correct glibc-2.2 symbol version one day, we'll just drop this on the floor (maybe after 2006.1) """ Think 'one day' has finally arrived? :)
Created attachment 255255 [details, diff] rebased 6610_all_glibc-r10k-workarounds.patch I rebased 6610_all_glibc-r10k-workarounds.patch on the current code. It was completely straightforward, but I don't have an R10k system to test on. Not compile tested, but if I fucked something up in that patch I shouldn't be allowed to use a text editor.
Created attachment 255259 [details, diff] rebased 6610_all_glibc-r10k-workarounds.patch This time with an #ifdef _MIPS_ARCH_R10000 so that we don't unnecessarily penalize other processors. Also, I've got a getpagesize patch to send before you add these to the -patches tarball.
(In reply to comment #4) > Created an attachment (id=255259) [details] > rebased 6610_all_glibc-r10k-workarounds.patch > > This time with an #ifdef _MIPS_ARCH_R10000 so that we don't unnecessarily > penalize other processors. Crap, can't do preprocessor checks inside a macro like that. First patch it is.
i didnt rebase the r10k patch because they changed the arguments to other insns. i dont know if the nops are necessary after the changes in registers. as for alpha, the fpic patch will probably have to have its original problem found and seen if it still exists. the cache shape patch should be sent upstream, and i think the fdatasync fix should be fixed too.
Created attachment 256007 [details, diff] Fix mips' getpagesize Attached is a patch that fixes mips' getpagesize.c. It includes ia64's implementation, which is just a preprocessed generic implementation. For whatever reason, ia64's getpagesize.c isn't in the work/ directory after unpacking the ebuild, so it's just easier to delete MIPS' getpagesize.c to force it to use the generic implementation. Also, forget about the r10k work around patch. I don't have any SGI boxes, so I'm not going to bother trying to support them. There have to be tons of hurdles to clear before indigo2/r10k works, and this one patch isn't going to do it without someone putting in some leg work. So, the alpha patches and this mips patch should go in, and I'd say kill 6600_mips_librt-mips.patch. Thanks vapier.
that patch is dependent on my rewrite of the ia64/common getpagesize code
(In reply to comment #8) > that patch is dependent on my rewrite of the ia64/common getpagesize code Ah, right. Patch 0055 for my future self. The MIPS patch should still be good to go. There's not really any reason why MIPS shouldn't be using the generic implementation. I'm running a locally patched glibc and it appears to work fine.
ive added your getpagesize to glibc-2.12.1 ... i dont think there's anything else for us to do here