fix(httpfs): disable SO_REUSEPORT in the test http server (flaky headCount under ctest -j) - #75
Merged
Merged
Conversation
RepeatOpenDoesNotIssueSecondHead failed flakily in the ladybug repo CI (headCount()==0) under `ctest -j10`. The vendored httplib enables SO_REUSEPORT by default on Linux, which lets multiple sockets bind the same port and load-balances connections between them. ctest runs each gtest case in its own process, so two concurrent test processes both successfully bound 127.0.0.1:18123 and the kernel routed some requests to the other process's server - the HEAD-count assertion then observed requests that never reached the server it was checking (reproduced locally: two concurrent processes both bound 18123, and the failing process saw headCount 0 or 2). Override the server socket options to set only SO_REUSEADDR so a second bind of the same port fails with EADDRINUSE and the existing port scan falls through to a genuinely free port. 20 consecutive `ctest -j10` runs pass with this change; without it ~25% of runs failed.
adsharma
force-pushed
the
fix/880-flaky-head-count-test
branch
from
September 1, 2026 16:59
f33f6f6 to
d115950
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Follow-up to #73 (the #880 fix).
HttpFileSystemTest.RepeatOpenDoesNotIssueSecondHeadfailed flakily in the ladybug repo CI withheadCount()==0underctest -j10.Root cause
The vendored httplib enables SO_REUSEPORT by default on Linux (
default_socket_optionsinthird_party/httplib/httplib.h), which allows multiple sockets to bind the same address:port and makes the kernel load-balance connections between them.gtest_discover_testsruns each gtest case in its own process, so withctest -j10the two httpfs test processes start a few ms apart and both successfully bound127.0.0.1:18123. Incoming connections were then distributed between the two servers by the kernel, so the process running the HEAD-count test could have some or all of its requests served by the other process's server (both serve the same fixture directory). Instrumented local repro showed exactly that: two processes both reportingbound port=18123, and the failing process observingheadCount0 or 2.Fix
LocalHttpServernow overrides the server socket options to set onlySO_REUSEADDR, so a second bind of the same port fails withEADDRINUSEand the existing port-scan loop picks a genuinely free port. Each test process then owns its port exclusively.Verification
ctest -j10runs fail locally (2-3 of 10).ctest -j10runs, 0 failures.httpfs_xetfs_testkeep passing.