russ@wpg.com (Russell Lawrence) (05/02/91)
In comp.lang.perl, article <3127@sparko.gwu.edu>, Randy McCaskill writes: rm> I had the same problem so I compiled it using Larry's malloc. It rm> passed the dbm test this way. It didn't really fix the problem, only rm> avoided it. If anyone knows of a better way, please post it. In article <=A_+TWB@smurf.sub.org>, Matthias Urlichs writes: mu> The problem is that some mallocs return NULL when you mu> allocate a zero-byte memory block. While that seems to be mu> legal behavior, some dbm libraries can't cope with this. mu> mu> I don't have a good fix either. (However, the Perl malloc mu> is faster anyway, at least on my machine.) I compiled perl4.003 without any immediate hitches on my system 5.2, 68020, multibus I machine using gcc with the -fpcc-struct-return and -traditional flags but I got "Out of Memory!" messages at the beginning of the dbm test suite until I switched to Larry's malloc. Now I'm getting a bus error message and a core dump on dbm test #2. A stack trace with adb gave me the following: _memcpy(6ac04)+1a _str_nset(4d184)+44 _str_make(77000000,41768)+38 _do_each(45de4)+7c _eval(45dc4)+27ca _eval(457c4)+290 _cmd_exec(41ec4)+10a4 _main(2,7fedf4,7fee00,2)+18e0 Not being expert with adb, the only thing I can gather from this is that we're introducing a stray pointer somewhere in the do_each function. Any suggestions? -- Russell Lawrence, WP Group, New Orleans (504) 443-5000 russ@wpg.com uunet!wpg!russ