GCC-3.0.1 can't compile Glibc-2.2.4

Jakub Jelinek jakub@redhat.com
Mon Sep 24 10:02:00 GMT 2001


On Mon, Sep 24, 2001 at 09:52:00AM -0700, H . J . Lu wrote:
> Jakub's change is almost ok to me. My only objection is dlopening 
> libgcc_s.so.1. I don't think it should be included unless it has
> been verified to work as intended. But it seems noone has tested if
> it works when the ABI of libgcc_s.so.1 changes, which is the main
> purpose of dlopening libgcc_s.so.1, and someone believes the ABI change

Nope, the main purpose of dlopening libgcc_s.so.1 is to catch when
unwind-dw2.c's internals (as opposed to the ABI of the functions which glibc
calls) change, which is likely to happen (think about adding new DWARF2.*
opcodes etc. - e.g. current unwind-dw2.c does not handle opcodes like
DW_OP_calli, DW_OP_push_object, DW_OP_piece if I name a few; current unwind-dw2.c
will abort() on them, libgcc_s.so from say gcc 3.2 might handle them and
that gcc might emit them).

> of libgcc_s.so.1 which dlopening libgcc_s.so.1 is trying to address is
> very unlikely to happen. Given those, I suggest
> 
> 1. Check in Jakub's change, excluding dlopening libgcc_s.so.1. Do a
> proper dlopening libgcc_s.so.1 later when necessary. Or
> 2. Check if dlopening libgcc_s.so.1 works, then check in the patch
> with a working dlopening libgcc_s.so.1 scheme.

	Jakub



More information about the Libc-alpha mailing list