More CI fixes#1854
Open
azat wants to merge 12 commits intolibevent:masterfrom
Open
Conversation
azat
force-pushed
the
more-ci-fixes
branch
2 times, most recently
from
9558c04 to
2f9bc49
Compare
The stddev tolerance of 100 was too tight for CI environments where scheduling variance naturally exceeds this threshold (observed 104.67). Increase to 150 to reduce flaky failures.
On macOS, the test setup calls thread_policy_set() to move threads to realtime scheduling class. If this fails (e.g. after fork() where Mach ports may be stale), the process called exit(1), killing the forked child and causing "[Lost connection!]" failures in CI. The realtime scheduling is a best-effort optimization for the test suite. Just warn and continue instead of aborting.
This test is flaky on Windows where socket closure detection timing differs from Unix, causing assertion failures on req and connection cleanup. Making it retriable allows CI to pass on retries.
Otherwise it is hard to understand what is going on here:
38/65 Test libevent#30: regress__POLL .......................***Failed 117.26 sec
FAILED
[Lost connection!]
[FAILED http/https_openssl_connection (0 retries)]
[msg] Nameserver 127.0.0.1:25962 has failed: request timed out.
[msg] All nameservers have failed
1/434 TESTS FAILED. (47 skipped)
On windows you can have different allowed ports for TCP and UDP, so we need to make sure that the selected port works for both. Refs: - docker/for-win#3171 - https://blog.deanosim.net/windows-10-winnat-and-why-your-programs-cant-listen-on-certain-ports/
CI:
FAIL D:\a\libevent\libevent\test\regress_bufferevent.c:1101: assert(labs(timeval_msec_diff(((&started_at)), ((&res1.write_timeout_at))) - (100)) <= 100): 140 vs 100bufferevent/bufferevent_timeout_filter_pair:
[FAILED bufferevent/bufferevent_timeout_filter_pair (0 retries)]
On windows: [RETRYING http/data_length_constraints (attempts left 2, delay 1 sec)] [RETRYING http/data_length_constraints (attempts left 1, delay 1 sec)] [RETRYING http/data_length_constraints (attempts left 0, delay 1 sec)]
Windows allocates ephemeral ports sequentially per protocol, and Hyper-V/WinNAT can exclude 200+ contiguous ports for TCP but not UDP (or vice versa). regress_pick_port() binds UDP first (port 0), then cross-checks TCP on the same port. When the UDP counter lands inside the TCP excluded range, all 5 sequential retries fail with WSAEACCES. Bump the retry count from 5 to 256 to reliably skip past any excluded range. Add diagnostic logging (port number + error) on bind failure for future troubleshooting.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.