Open nfi-hashicorp opened 1 year ago
Hello! That's correct, the -test.timeout
flag would only limit the time of each re-run. A --timeout
flag seems like a reasonable addition, but I would like to avoid flags that mirror existing go test
flags. Maybe --rerun-timeout
to make it clear that it's only relevant to reruns.
Another option might be to examine the args
passed to go test
, and if we notice a -test.timeout
or -timeout
flag, use that timeout automatically. That won't work with --raw-command
if the timeout is in the command and not the command line.
As a workaround the timeout
command on linux should also work pretty well:
timeout -s INT 100 gotestsum ...
This signal handler will forward the interrupt to go test
. With this approach we don't get the stack trace.
Another improvement would be to send SIGABRT
to go test
, which should give us a goroutine stack trace (like -test.timeout
would) instead of only interrupting the run. I opened #346 to add support for sending SIGABRT
.
You can use the
go test -timeout
flag, but AFAICT, that wouldn't take into account retries.