lang/clisp vs. gcc?

[Havard Eidnes <he%NetBSD.org@localhost>]

Hi,

bulk-building pkgsrc-current of a while ago on NetBSD/macppc 8.99.25,
I see the build stalls (apparently until CPU time limit is reached,
which I've configured for the build job) on lang/clisp:

load averages:  0.81,  0.81,  0.91;               up 107+08:22:43      18:21:53
52 processes: 1 runnable, 50 sleeping, 1 on CPU
CPU states:  0.0% user,  100% nice,  0.0% system,  0.0% interrupt,  0.0% idle
Memory: 186M Act, 136M Inact, 7568K Wired, 24M Exec, 272M File, 94M Free
Swap: 1300M Total, 132M Used, 1168M Free

  PID USERNAME PRI NICE   SIZE   RES STATE      TIME   WCPU    CPU COMMAND
14559 1138      23    4  5976K 1552K RUN       80:05 97.12% 97.12% lisp.run
    0 root     126    0     0K   28M pgdaemon 598:51  0.00%  0.00% [system]

(gdb) where
#0  0x01921120 in calloc ()
#1  0xfdec626c in ?? () from /usr/lib/libintl.so.1
#2  0xfdec63c8 in bindtextdomain () from /usr/lib/libintl.so.1
#3  0xfdec3688 in dcngettext () from /usr/lib/libintl.so.1
#4  0xfdec47fc in dgettext () from /usr/lib/libintl.so.1
#5  0x01926aa4 in main ()
(gdb) detach

I vaguely seem to recall that there's a chance this is the result of
gcc being too clever by half and doing weird things with calloc() /
malloc() optimization?  I have clisp suspected of (trying to?) define
one of these or both itself, leading to this rather tight CPU-bound
loop.

Do we have a recipe for working around situations such as this?

Regards,

- Håvard