cliff@unmvax.UUCP (04/08/85)
Lately I haven't been able to get mail to mcvax. This letter concerns
hack 1.0. On our SUN, wands of wishing cause core dumps when items are
misspelled. I found the problem to be related to NROFOBJECTS being defined
improperly. My quick patch was to change the following line (we are still
running hack 1.0).
436c436
< while(i <= NROFOBJECTS && (!let || objects[i].oc_olet == let)){
---
> while(objects[i].oc_name && (!let || objects[i].oc_olet == let)){
--Cliffaeb@mcvax.UUCP (Andries Brouwer) (04/10/85)
In article <774@unmvax.UUCP> cliff@unmvax.UUCP writes: > Lately I haven't been able to get mail to mcvax. This letter concerns > hack 1.0. On our SUN, wands of wishing cause core dumps when items are > misspelled. I found the problem to be related to NROFOBJECTS being defined > improperly. My quick patch was to change the following line (we are still > running hack 1.0). > > 436c436 > < while(i <= NROFOBJECTS && (!let || objects[i].oc_olet == let)){ > --- > > while(objects[i].oc_name && (!let || objects[i].oc_olet == let)){ Thanks for the bug report; in fact your letter reached me, but perhaps my reply failed to reach you. Anyway, I don't think there is anything wrong with NROFOBJECTS, the problem is rather that strcmp(p,q) is done where q may be 0. Your attempt at a fix only solves part of the problem - a core dump is still possible - while on the other hand it makes asking for certain objects impossible. I think the following would be better (the line numbers are from hack.objnam.c version 1.0.2 but the fix works equally well on earlier versions). # diff hack.objnam.c (new) hack.objnam.c (old) 498,501c498 < register char *zn = objects[i].oc_name; < < if(!zn) goto nxti; < if(an && strcmp(an, zn)) --- > if(an && strcmp(an, objects[i].oc_name)) 503c500 < if(dn && (!(zn = objects[i].oc_descr) || strcmp(dn, zn))) --- > if(dn && strcmp(dn, objects[i].oc_descr)) 505c502 < if(un && (!(zn = objects[i].oc_uname) || strcmp(un, zn))) --- > if(un && strcmp(un, objects[i].oc_uname))