On Mon, Aug 24, 2009 at 8:39 AM, Andrew Whitworth<[email protected]> wrote: > On Sun, Aug 23, 2009 at 10:45 PM, Austin > Hastings<[email protected]> wrote: >> At any rate, svn-bisect tells me that the problem comes from r40628, which >> naturally has a whole bunch of files associated with it. (It's never the >> one-line change that gets me... :( ) > > For those who aren't aware, r40628 was the merger of the auto_attrs > branch. That branch affected the way Parrot_*_attributes structures > are allocated for PMCs. Instead of manually allocating these > structures in VTABLE_init and then having to manually free them again > in VTABLE_destroy, the PMC allocator now reads the necessary size from > the VTABLE structure and allocates (and deallocates) memory > automatically.
Verified; see https://trac.parrot.org/parrot/ticket/945 While the tcl tests didn't start running out of memory until some commits after this, this commit is probably the culprit. > I was looking through the code yesterday and found a place in the GC > where I think we are not freeing the attributes structures like we > should. This would mean we are leaking memory after a GC run. If I am > right about the problem the fix should be a relatively easy one. I'll > try to put a patch together soon, but I can't do any local testing > really until I get home from work. > > --Andrew Whitworth > _______________________________________________ > http://lists.parrot.org/mailman/listinfo/parrot-dev > -- Will "Coke" Coleda _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
