tests: tentative fixes for sporadic host and kvm failures #3434
Conversation
| func waitOrFail(t *testing.T, child *gexpect.ExpectSubprocess, expectedStatus int) { | ||
| bufOut := []string{} | ||
| ttyIn, ttyOut := child.AsyncInteractChannels() | ||
| close(ttyIn) |
There was a problem hiding this comment.
👍 Can we leave a TODO to change gexpect.AsyncInteractChannels to accept channels rather than returning them?
There was a problem hiding this comment.
Do you mean opening a ticket to gexpect or a just a reminder for us here?
There was a problem hiding this comment.
I suggest just a TODO here, I'll set up a PR in gexpect and some documentation around the usage patterns for these channels.
| bufOut := []string{} | ||
| ttyIn, ttyOut := child.AsyncInteractChannels() | ||
| close(ttyIn) | ||
| for line := range ttyOut { |
There was a problem hiding this comment.
(I guess) this relies on the fact that the ttyOut chan will be closed when a real err happens or io.EOF in https://github.com/ThomasRooney/gexpect/blob/8e96656e8fb794d716ee75b5a8cecbc15cde4393/gexpect.go#L167. I think we should document this here and in gexpect.
There was a problem hiding this comment.
It does, I'll note this down.
lucab
changed the title
lucab
changed the title
This commit fixes waitOrFail to drain output from the child, in order to avoid a deadlock situation where the children is blocked on a write toward the parent, while the parent is blocked waiting for the children.
This raises the pod-stop timeout from 5 to 15 seconds. Sporadic test failures on KVM flavor may be due to a too strict limit, so we just 3x here to make sure failures are not due to bad timing.
lucab
force-pushed
the
to-upstream/tests-fixes
branch
from
fd8ee7a
to
46e9c15
Compare
|
Rebased, PTAL. |
This PR tentatively fixes the following flakes: