On 2012-12-29, Volker Braun <vbraun.n...@gmail.com> wrote: > ------=_Part_750_16846180.1356813944502 > Content-Type: text/plain; charset=ISO-8859-1 > > Fix is at http://trac.sagemath.org/13880 > > Also, I don't see how a longjmp is going to fix that. Ideally, GAP would > rewrite their allocator to not require a single contiguous chunk of memory.
GAP uses mmap to get/extend memory. On Linux, mmap has a MAP_NORESERVE parameter which tells the kernel whether or not to reserve the swap. Granted, if MAP_NORESERVE is set then one can get a segfault - but segfaults can be caught with longjumps. On OSX and other Sage-fully-supported (tm) platforms MAP_NORESERVE might not be available, but this is a different story. > It would be fine to have a fixed-size master pointer area and a > reallocatable (or even non-contiguous) workspace for memory bags. But I > don't think we should patch that in ourselves. Right. In fact GAP people are working on switching over to Boehm GC, although I don't have any timeframe for this change to happen. Dima -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To post to this group, send email to sage-devel@googlegroups.com. To unsubscribe from this group, send email to sage-devel+unsubscr...@googlegroups.com. Visit this group at http://groups.google.com/group/sage-devel?hl=en.
