Am Donnerstag, 10. Mai 2007 23:07 schrieb Andrew Dougherty: > On Wed, 9 May 2007, Andy Dougherty wrote: > > Ok, I've got it compiling, at least. I'll let it run tests overnight and > > submit a cleaned-up patch tomorrow. I've gone back and reviewed the > > alignment issue and think I have implemented it sanely. > > Hmm. Rats. My initial version required more invasive changes elsewhere > (esp. in string.c and dod.c) than I'm comfortable with. It still needs > more work.
Well, I saw some patches [1] floating by which did remove PObj_bufstart_external_FLAG. This flag was needed to sync the malloc-based allocators with parrots (and COWs) understanding of a 'not owned' / external buffer/string memory resource. COWed strings are not external to the malloc-based GCs, they just have an increased ref-count. I was never able to fix string allocation and COW in such a way that this flag would be obsolete with GC_IS_MALLOC. leo [1] well it seems ;) that I've committed r10582 but this patch obviously ignroed the GC_IS_MALLOC-cases. While having different allocators is nice, it's really hard to test & sync them with the default one.