Front page | perl.perl5.porters |
Postings from December 2003
On Sun Dec 21 22:57:34 2003, Nicholas Clark wrote: > On Sun, Dec 21, 2003 at 10:39:50PM +0000, Marty Pauley wrote: > > If you're feeling paranoid, a safer place to unOOK is much closer to the > > start of pp_subst, just before we take the first pointer to the inside > > of the PV. > > What do I do if I'm feeling really paranoid, and wondering if there are a > whole class of bugs like these? I think there might be a few more like this. We could also get similar problems when we grow the sv and cause the pv to change altogether. We should get a segmentation fault in that case. The thing that worried me the most about this one was that there were cases that hit the bug but still 'worked': when the offset was larger than the remainder of the string, the move wouldn't overwrite the old part of the string. > [taking pointers to the insides of SVs that are then moved around because > of upgrading. Hack the core to create all scalars as OOK, and see which > tests fail?] That might be a good start. :-) If we had a version of Move that filled the source with rubbish, we might be able to find some more bugs. And if we changed sv_backoff and sv_grow to release memory when they could, we might get some nice segmentation faults to indicate bugs. -- Marty