/cc @vdemeester @thaJeztah @anusha-ragunathan PTAL 😄

Codecov Report

Merging #37076 into master will decrease coverage by 0.25%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #37076      +/-   ##
==========================================
- Coverage   35.34%   35.09%   -0.26%     
==========================================
  Files         615      615              
  Lines       45818    45818              
==========================================
- Hits        16196    16080     -116     
- Misses      27470    27628     +158     
+ Partials     2152     2110      -42

Do we know what caused the test to fail in the first place?

Do we know what caused the test to fail in the first place?

It's hard to say as hard to reproduce, according to last experience, this kind of issue most probably is a synchronization issue, but since we have a 3rd party pty code, the pty.Start() is also a potential culprit.
Time consuming to narrow down the real cause, IMO we can focus on the test case itself if it can check the original regression or not?

Just take a look at the Start() method in vendor/github.com/kr/pty/run.go, seems it's the caller's responsibility to Close() the pty *os.File, so I guess the root cause maybe: we copy data from an opened file which will never be closed, from the file system perspective, this will probably result in data synchronization issue.

I suggest to add below code snippet in TestExecInteractiveStdinClose:

go func() {
  io.Copy(b, p)
  p.Close()
 } ()

Do we need to squash this change?

@arm64b interesting… we should try that 👼 (in another PR, this one being merged already)

+1 on trying #37076 (comment)

Follow-up done in #37086