Add more verifications in EventPipe rundownvalidation test. #89623
Conversation
Subtle changes like: dotnet@2f2aaae Knocked out a couple of rundown events on Mono that was detected and fixed in dotnet#88634. It would been ideal to catch this issue as part of CI, but turns out that our existing rundownvalidation test only tests a couple of rundown events, and not any of the module/assembly events that was affected by above change. This commit adds validation that the following additional rundown events are included in the rundown event stream: RuntimeStart MethodDCStopInit MethodDCStopComplete LoaderModuleDCStop LoaderDomainModuleDCStop LoadAssemblyDCStop LoadAppDomainDCStop
lateralusX
added
area-Diagnostics-mono
area-Diagnostics-coreclr
and removed
area-Tracing-coreclr
labels
|
Tagging subscribers to this area: @tommcdon Issue DetailsSubtle changes done in: like a change where one of the utf8->utf16 conversion routines changed behavior when passing in a string that is "" and len 0. In previous implementation it would return a new string "", but in new update change, it starts to return NULL. That in turn would cause some fields in some EventPipe events on Mono to fail serialize (using "" if value is not present), and that knocked out a couple of rundown events on Mono that was detected and fixed in #88634. It would been ideal to catch this issue as part of CI, but turns out that our existing rundownvalidation test only tests a couple of rundown events, and not any of the module/assembly events that was affected by above change. This commit adds validation that the following additional rundown events are included in the rundown event stream: This would at least make sure the majority of important rundown events are present.
|
|
@lateralusX, I didn't realize this difference in eventpipe. Thanks for the fix! Now that both coreclr and mono ep implementations end up calling the same |
ghost
locked as resolved and limited conversation to collaborators
Subtle changes done in:
#85558
like a change where one of the utf8->utf16 conversion routines changed behavior when passing in a string that is "" and len 0. In previous implementation it would return a new string "", but in new update change, it starts to return NULL. That in turn would cause some fields in some EventPipe events on Mono to fail serialize (using "" if value is not present), and that knocked out a couple of rundown events on Mono that was detected and fixed in #88634.
It would been ideal to catch this issue as part of CI, but turns out that our existing rundownvalidation test only tests a couple of rundown events, and not any of the module/assembly events that was affected by above change.
This commit adds validation that the following additional rundown events are included in the rundown event stream:
This would at least make sure the majority of important rundown events are present.