tests/runtests.pl: use default post command delay of 100ms #7833
Conversation
|
It will be interesting to know the result of this, however I don't think it's a good idea to actually put it in. |
mback2k
force-pushed
the
win-tests-delay
branch
2 times, most recently
from
4caaecb
to
da23eb1
Compare
|
It seems like this is actually helping, but on the other side some test runs are now timing out. I will have to dive deeper into the timings. The only remaining errors are either unrelated or those server unexpectedly alive errors which I am also trying to work on. I also would like to avoid putting this in, but after countless tries of adding/forcing flushing behavior between the Perl/Bash/Binary process chains in our test suite, it still seems like Windows is not behaving correctly about this. Unfortunately my observation after all these months is that flakiness on Windows is caused by two things a) missing lines in process output and b) test servers being unexpectedly alive and therefore not setup correctly. |
mback2k
force-pushed
the
win-tests-delay
branch
2 times, most recently
from
e3a0db6
to
a358780
Compare
Although flushing the output buffers reduced the flakiness of the curl test suite on Windows already, some tests were still randomly failing. This is another attempt to fix it. Follow up to curl#7530
mback2k
force-pushed
the
win-tests-delay
branch
from
a358780
to
a6ebe28
Compare
On Windows data can be lost in buffers in case of abnormal program termination, especially in process chains as seen due to flaky tests. Therefore flushing all buffers manually should avoid this data loss. In the curl tool we play the safe game by only flushing write buffers, but in the testsuite where we manage all buffers, we flush everything. This should drastically reduce the CI and testsuite flakiness. Supersedes curl#7833 and curl#6064
On Windows data can be lost in buffers in case of abnormal program termination, especially in process chains as seen due to flaky tests. Therefore flushing all buffers manually should avoid this data loss. In the curl tool we play the safe game by only flushing write buffers, but in the testsuite where we manage all buffers, we flush everything. This should drastically reduce Windows CI and testsuite flakiness. Reviewed-by: Daniel Stenberg Supersedes #7833 and #6064 Closes #8516
|
Closing in favor of #8516. |
Although flushing the output buffers reduced the flakiness
of the curl test suite on Windows already, some tests were
still randomly failing. This is another attempt to fix it.
Follow up to #7530