5.1.2 test failures on alphaev56-dec-osf4.0g

[Daniel Richard G.]

On Tue, 2013 Sep 17 23:32+0200, Niels Möller wrote:
> 
> What about more recent versions of gcc?

I ultimately want to build GCC 4.5.4, but it depends on GMP :-)

> > FAIL: t-invert
> > FAIL: t-div
> 
> I'd suggest investigating these two failures first.

I recompiled with --enable-assert and --enable-alloca=debug, and invoked
t-invert in a debugger. When I saw this...

    Program received signal SIGSEGV, Segmentation fault.
    __gmpn_invert_limb () at tmp-invert_limb.s:61
    61      tmp-invert_limb.s: No such file or directory.

...I got deja vu:

    http://gmplib.org/list-archives/gmp-bugs/2010-September/002043.html

Here are the relevant bits of GDB session, re-run with the assembly
files still in place:

    Starting program: /tmp/gmp-test/tests/mpn/t-invert
    [...]
    Program received signal SIGSEGV, Segmentation fault.
    __gmpn_invert_limb () at tmp-invert_limb.s:61
    61              lda   $2,  approx_tab

    Starting program: /tmp/gmp-test/tests/mpn/t-div
    [...]
    Program received signal SIGSEGV, Segmentation fault.
    __gmpn_invert_limb () at tmp-invert_limb.s:61
    61              lda   $2,  approx_tab

For easy reference, here is the apropos excerpt from tmp-invert_limb.s:

__gmpn_invert_limb:
        ldgp    $29,0($27)
        .frame $30,0,$26
        .prologue 1
        lda   $2,  approx_tab  <--- SIGSEGV
        srl     $16, 54, $1
        srl     $16, 24, $4
        and     $16, 1, $5
        bic     $1, 1, $7

This Alpha system does not seem to like that instruction.


--Daniel


-- 
Daniel Richard G. || skunk at iSKUNK.ORG
My ASCII-art .sig got a bad case of Times New Roman.