Bug#231747: libssl0.9.7: doesn't use /usr/lib/v8 on sparc32 when possible
So theres this bug(231747), which makes things like ssh crawl on a ss20...
if you run say,
(%:/tmp/glibc-2.3.2.ds1)- grep -r HWCAP_SPARC_ .
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_FLUSH 1 /* The cpu supports flush insn. */
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_STBAR 2
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_SWAP 4
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_MULDIV 8
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_V9 16 /* The cpu is v9, so v8plus is ok. */
./build-tree/glibc-2.3.2/elf/elf.h:#define HWCAP_SPARC_ULTRA3 32
./build-tree/glibc-2.3.2/sysdeps/unix/sysv/linux/sparc/sparc32/dl-procinfo.h:#define HWCAP_IMPORTANT (HWCAP_SPARC_V9|HWCAP_SPARC_ULTRA3)
./build-tree/glibc-2.3.2/sysdeps/unix/sysv/linux/sparc/sparc64/dl-procinfo.h:#define HWCAP_IMPORTANT (HWCAP_SPARC_ULTRA3)
./build-tree/glibc-2.3.2/sysdeps/sparc/sparc32/dl-machine.h:#define LD_SO_PRELOAD ((GL(dl_hwcap) & HWCAP_SPARC_V9) \
./build-tree/glibc-2.3.2/sysdeps/sparc/sparc32/dl-machine.h: return GL(dl_hwcap) & GL(dl_hwcap_mask) & HWCAP_SPARC_V9;
./build-tree/glibc-2.3.2/sysdeps/sparc/sparc32/dl-machine.h: return GL(dl_hwcap) & HWCAP_SPARC_V9;
./build-tree/glibc-2.3.2/sysdeps/sparc/sparc32/dl-machine.h: do_flush = GL(dl_hwcap) & HWCAP_SPARC_FLUSH;
./build-tree/glibc-2.3.2/sysdeps/sparc/sparc32/dl-machine.h: int do_flush = GL(dl_hwcap) & HWCAP_SPARC_FLUSH;
./build-tree/glibc-2.3.2/ChangeLog.7: * elf/elf.h (HWCAP_SPARC_*): New definitions.
./build-tree/glibc-2.3.2/ChangeLog.8: (HWCAP_SPARC_V9): Add.
./build-tree/glibc-2.3.2/ChangeLog.12: * elf/elf.h (HWCAP_SPARC_ULTRA3): Define it.
it looks like nothing will ever set HWCAP_SPARC_MULDIV, which I think is
v8...
does there just need to be a few more lines in dl-machine.h for v8?
This also might have something to do with it:
(%:/tmp/glibc-2.3.2.ds1)- egrep -r sparcv._LIBDIR .
./debian/sysdeps/sparc.mk:sparcv9_LIBDIR = /v9
though the libs do get put in /usr/lib/v8 just fine...
I'd try to fix this myself, but I have a feeling someone knows exactly
what needs to be done :-)
--
-Justin
Reply to: