"Adam J. Richter" wrote:
>
> Garteth Hughes's "Pentium III FXSR, SSE support" that was
> added to linux-2.4.0-test1ac8/include/asm-i386/sigcontext.h causes
> glibc to fail to compile.
>
> The glibc code that it breaks is lines 142 through 205 of
> glibc-2.1.3/sysdeps/unix/sysv/linux/i386/register-dump.h,
> which is come code in the register_dump debugging routine that
> attempts to print the floating point state. dump_registers is called
> by catch_segfault in glic-2.1.3/sysdeps/generic/segfault.c, which
> is called only when this facility is enabled and the program gets a
> segmentation fault.
>
> I have temporarily commented out the lines that do not
> compile, since I think we do not currently use this facility, but
> a better fix should be possible, at least through some glibc
> configure.in test and some #ifdefs, since there is already support in
> the modified sigcontext structure for determining which format of
> fp data is being used (although putting "magic" after the union
> instead of before it will limit future extensibility, although
> perhaps that was necessary for current downward binary compatability).
>
> What I would like to know is how fixed is the new
> struct _fpstate, and how certain is it that these changes are going
> into the stock linux kernel when Linus gets back from vacaction.
> If this is basically the new official format of struct _fpstate,
> I would be willing to write a fix for glibc if nobody else wants to.
I'm submitting a patch that fixes this today to both the kernel and
glibc. Ulrich and I have been discussing this code for a couple of days
and we've agreed on the best format for things.
-- Gareth
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:14 EST