Fix data race in FST Segment Reads #938
Conversation
Codecov Report
@@ Coverage Diff @@
## master #938 +/- ##
==========================================
- Coverage 77.98% 77.96% -0.03%
==========================================
Files 411 411
Lines 34407 34406 -1
==========================================
- Hits 26834 26823 -11
- Misses 5730 5737 +7
- Partials 1843 1846 +3
Continue to review full report at Codecov.
|
| ) | ||
| wg.Add(2) | ||
|
|
||
| for i := 0; i < 2; i++ { |
There was a problem hiding this comment.
Is it worth adding some more parallelism to more likely hit the edge case?
There was a problem hiding this comment.
Nah, even this version with 2 routines fails pretty consistently within 1-2 of the sub-tests running (of the 100 prop test runs per test execution). No point adding more, it'll only serve to slow down CI
| errLock sync.Mutex | ||
| matchErr error | ||
| ) | ||
| wg.Add(2) |
There was a problem hiding this comment.
super nit: Perhaps put a number up front which you can pass to both wg.Add(...) and the for loop so that they always match (easier to turn up/down parallelism).
There was a problem hiding this comment.
sure, opened PR #940
H/T @sboschman for debugging and suggesting the issue.
Example failure of added test without fix - https://pastebin.com/mtBbNUr6