|
From: Nicholas N. <nj...@cs...> - 2007-07-17 09:44:09
|
On Mon, 16 Jul 2007, Christoph Bartoschek wrote: > Open questions: > > 1. What could cause the leak tracer regressions? Those regression tests are sensitive and sometimes fail anyway. So it's hard to tell if it's a true regression -- sometimes you can tell by seeing how different the expected results are from the actual results. > 2. What to do with the need_discard parameter of the VG_(am_munmap_client)() > function? > > Maybe you could look at the code and see some mistakes. I currently do not > think that this is something for the mainline valgrind. But maybe others also > want to run big programs with lots of allocations with valgrind and are > interessted in a clean solution. I hope to take a look at your code soon, but my time is limited. I know Julian is also interested in your changes but he is very busy with non-Valgrind things right now. So it may take a while for us to act on this, but we do plan to. Thanks for your hard work in hunting down these problems! You are definitely on the bleeding edge, running such large programs, and I hope you can get some useful information out of Valgrind now :) Nick |