Conversation
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Finally tracked down the random crashes to an issue in Catch2. catchorg/Catch2#3031 to fix |
|
Going to start documenting failures/hangs here so I don't lose track of them. Here's the latest hang: |
|
Another hang: |
10baf45 to
fd4f5c9
Compare
| } | ||
|
|
||
| // NOTE: Requires 's_lock' to be held | ||
| apartment_variable_storage* ensure_current_apartment_variables() |
There was a problem hiding this comment.
If it requires the lock to be held should it take one of those WIL lock holder parameters to try and enforce that?
There was a problem hiding this comment.
All changes here were just to move internal functions to be protected instead of public.
| { | ||
| _doneHangingHandle.SetEvent(); | ||
| } | ||
| CWMO_DISPATCH_CALLS | CWMO_DISPATCH_WINDOW_MESSAGES, INFINITE, ARRAYSIZE(handles), handles, &index)); |
There was a problem hiding this comment.
Changing a 10 second timeout to infinite isn't going to help with tests hanging :|
There was a problem hiding this comment.
Maybe not, but I'm trying to remove non-determinism as much as I can, especially since CI machines are unpredictable. I've added some "hang detection" logic to terminate tests after 10 min w/ callstacks, so INFINITE waits aren't that bad anymore.
There was a problem hiding this comment.
That said, I haven't (yet) turned all waits INFINITE, but maybe I should...
| wil::shared_event doneHangingHandle; | ||
| doneHangingHandle.create(); | ||
| // There is a race between cancellation and the underlying implementation getting called such that it's possible for the call | ||
| // to get marked as cancelled and then the marshalling code sees that the call is cancelled and never invoke the |
There was a problem hiding this comment.
FYI - this draft is just to get CI runs more easily (yeah I could technically schedule on my branch, but I won't get runs on push). I'm likely to delete this PR & re-issue a new one once everything is ready
af0d282 to
9fbe99c
Compare
Tests pretty reliably hang and time out when running in CI. Trying to track down the issue