Closed jmid closed 10 months ago
Cygwin-wise, the 4 merged workflows completed 69m, 76m, 79m, and 84m, thus confirming that merging should be OK. We may even consider lowering the time out at some point...
With the removal of the bytes
dependency in qcheck-core.0.21.3
we had a record low 12m run on Linux testing 5.1 - and 12 complete GA runs of <=20m! :tada:
CI summary:
Lin Out_channel test with Domain
#412threadomain
#404threadomain
#203 Internal error
Out of 59 workflows 4 failed with 3 genuine issues and 1 CI issue.
I updated the PR according to my suggestion :-)
Of the 3 Cygwin runs 3 completed in 80m, 87m, 127m and 1 timed out after 6h because of a deadlock.
CI summary:
Dynlink
test #307Lin Dynlink
test #307 Fatal error: Failed to create domain
during domain_spawntree
#354thread_joingraph
#353thread_createtree
#353thread_joingraph
and the Lin Stack
test triggered assertion failure in memprof.c #422Lin Queue
, Lin Bytes
, Lin.Internal ref int test with Thread
, and Lin.Internal CList int test with Thread
all triggered assertion failures in domain.c #404threadomain
, Lin Queue
, Lin Stack
, Lin Bytes
, and lin_internal_tests_thread_conclist
all triggered an assertion failure in memprof.c #422stm_tests_thread_ref
and lin_internal_tests_thread_ref
triggered an assertion failure in domain.c #404threadomain
#203domain_spawntree
#354domain_spawntree
#354Internal error
Internal error
Lin Out_channel test with Domain
counterexample #401Out of 59 workflows 13 failed, with 10 triggering genuine errors, 1 false alarm, and 2 internal CI failures
CI summary for merge to main
:
domain_spawntree
with Failed to create domain
#354thread_createtree - with Atomic
#425Out of 36 workflows 5 failed with 3 genuine issues and 2 CI setup-related failures
This PR merges the two GitHub actions workflow parts for Cygwin into a single one.
Initially, the split was done out of necessity in https://github.com/ocaml-multicore/multicoretests/pull/305 and https://github.com/ocaml-multicore/multicoretests/pull/313 as the Cygwin port initially was too slow to complete within the 6h timeout!
Since then we have seen general OCaml (GC) improvements, meaning that the 2-part Cygwin CI workflow would complete in about 2x45m, and thus opening up for a workflow merge incl. saving time on storing and restoring the state.
Finally I offer two data points from a trial run of this branch:
5.1.1~rc1
completed in 1h16mtrunk
completed in 1h11m