Simon, for debugging and correctness check purposes it would be great to have a hook to run an optional garbage collection (or anything else) before each access to a weak dictionary cache in Sage. Did you ever think about something like that? Right now one can abuse the python debugger hook to continuously collect garbage but that slows Sage down to a crawl.
On Friday, May 4, 2012 12:13:27 PM UTC-4, Simon King wrote: > > Hi Willem Jan, > > On 4 Mai, 17:37, Willem Jan Palenstijn <w...@usecode.org> wrote: > > > Anyway, I still wonder: Can't one use valgrind to detect those things? > > > > > I tried > > > valgrind --log-file=... ./sage -t > devel/sage/sage/crypto/mq/mpolynomialsystem.py > > > and then "glibc detected" did show up on screen. But after > interrupting > > > with Ctrl-c, the log file wasn't exactly helpful (just a handful of > > > lines). > > > > Did you try ./sage -t -valgrind devel/sage/sage/.... ? > > No, since I don't have the optional valgrind spkg installed - yet... > > > You probably need either the valgrind optional spkg or just the > sage.supp from > > it in local/lib/valgrind/sage.supp. > > > > Running "valgrind ./sage -t ...." will just debug the shell executing > the sage > > script, not sage itself. (--trace-children=yes should get around that, > by the > > way, but ./sage -t -valgrind would be preferable if that works.) > > I'll certainly try! > > BTW, I made some progress, in the sense that changing a weak > dictionary into a strong dictionary fixed that test - however, not all > is good: Another test fails with exactly the same message, and passes > if the strong dictionary is reverted into a weak dictionary! > > Anyway, it seems to confirm that the problem comes from garbage > collection. > > Best regards, > Simon > -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
