genattrtab cores on bootstrap was: gcc-ss-20000228 is now available

[Robert Lipe <robertlipe at usa dot net>]

Benjamin Scherrey wrote:
> Hmmm... this build fails for me using gcc-2.95.2 under Intel Linux
> 2.2.13 (RedHat 6.1) and binutils 2.9.1. I do a ./configure
> --enable-shared;make bootstrap and it comes down to:
> 
> ./genattrtab ./config/i386/i386.md > tmp-attrtab.c
> make[2]: *** [s-attrtab] Error 139

Are you peraps seeing a fault like this?

0x804ed41 in clear_struct_flag (x=0x0) at /play/egcs/gcc/genattrtab.c:4030
4030      MEM_IN_STRUCT_P (x) = 0;

(gdb) where
#0  0x804ed41 in clear_struct_flag (x=0x0) at /play/egcs/gcc/genattrtab.c:4030
#1  0x804e040 in optimize_attrs () at /play/egcs/gcc/genattrtab.c:3495

(gdb) up
#1  0x804e040 in optimize_attrs () at /play/egcs/gcc/genattrtab.c:3495
3495            clear_struct_flag (iv->av->value);
(gdb) print iv->av
$2 = (struct attr_value *) 0x14a
(gdb) print iv->av->value
$3 = (struct rtx_def *) 0x0
(gdb) print *iv
$4 = {av = 0x14a, ie = 0x819fff0, attr = 0x14b, next = 0x14b}

My guess is that we're walking past the end of insn_code_values[].


This is one of my three current mysterious failure modes.  This is the
failure I've seen for weeks on my UDK-targeted on OpenServer build....

RJL