Clean up tar file at the end of WriteEntry_LongFileSizeAsync test#79907
Clean up tar file at the end of WriteEntry_LongFileSizeAsync test#79907akoeplinger wants to merge 1 commit intodotnet:mainfrom
Conversation
Otherwise multiple tar files could stick around for some time, contributing to issues like dotnet#77012
| Assert.Equal(size, dataStream.Position); | ||
|
|
||
| Assert.Null(reader.GetNextEntry()); | ||
| tarFile.Close(); |
There was a problem hiding this comment.
I would expect #79899 to solve this problem for good. If that is not the case, we need to investigate it and ensure that Dispose does the right thing.
There was a problem hiding this comment.
it didn't solve the issue for me, at least on Mono. I haven't checked on CoreCLR (GC behavior can be quite different between the runtimes)
There was a problem hiding this comment.
ah I see, I missed that there's a using when creating the TarReader so we should expect it getting disposed at the end of the test.
There was a problem hiding this comment.
hmm, interesting: without the change in this PR it looks like the file does get deleted on disk however the process still keeps it open at least according to Activity Monitor on Mac:
and it also still counts against the Available disk space in df -h.
There was a problem hiding this comment.
seems to be specific to the Async test
There was a problem hiding this comment.
ahhh the fix in #79899 didn't get applied to the DisposeAsync() method 😆
There was a problem hiding this comment.
ok and another problem is that the WrappedStream isn't disposing the inner _baseStream. I'll be away for a few hours now, but I can send a PR later.
There was a problem hiding this comment.
akoeplinger
added
the
NO-MERGE
ghost
locked as resolved and limited conversation to collaborators
3 participants
Otherwise multiple tar files could stick around for some time, contributing to issues like #77012
With this I've never seen disk usage go below the 8GB that a single test run takes.