Repository navigation
Fix memory leaks in OMRCodeCacheManager.cpp - #7627
Conversation
discovered with valgrind working on JB2 Signed-off-by: Mark Stoodley <mstoodle@ca.ibm.com>
|
@mpirvu any thoughts on the safety of destroying the cache list mutex and repository monitor as part of destroying the CodeCacheManager? I seem to recall we've run into shutdown issues around cleaning up the code cache manager in the past. Do you recommend I find a way to isolate OpenJ9 from this change? |
|
As far as I can tell we don't destroy the manager in OpenJ9, so this change should be safe. |
|
I don't believe this will break OpenJ9 either |
|
@dsouzai could I get a review from you here, please, when you have a moment? |
dsouzai
left a comment
There was a problem hiding this comment.
Looks fine to me both from an OMR pov and an OpenJ9 pov; for the latter, because the compilation and sampler threads are stopped before destroying the code cache manager, no one needs to acquire those monitors at this point.
|
jenkins build all |
1 similar comment
|
jenkins build all |
|
ppcle failure due to #6571; zos nodes are offline; riscv build was aborted because the jenkins CI went down. Don't see a point in restarting riscv as it takes a long time and this change is common code. Merging. |
discovered with valgrind working on JB2, may be worth some openj9 tests before merging so I'll leave as WIP for now.