T4334 breaks CI on master (stat Decrease from x86_64-linux-deb9-debug)
The T4334
test is wreaking havoc on CI, both on master
(!) and elsewhere. Here is an example:
runtime/peak_megabytes_allocated Decrease from x86_64-linux-deb9-debug baseline @ HEAD~7:
Expected T4334 (nonmoving_thr) runtime/peak_megabytes_allocated: 112.0 +/-2%
Lower bound T4334 (nonmoving_thr) runtime/peak_megabytes_allocated: 109
Upper bound T4334 (nonmoving_thr) runtime/peak_megabytes_allocated: 115
Actual T4334 (nonmoving_thr) runtime/peak_megabytes_allocated: 105
Deviation T4334 (nonmoving_thr) runtime/peak_megabytes_allocated: -6.2 %
*** unexpected stat test failure for T4334(nonmoving_thr)
...
Unexpected stat failures:
perf/space_leaks/T4334.run T4334 [stat Decrease from x86_64-linux-deb9-debug baseline @ HEAD~7] (nonmoving_thr)
This has been independently observed in the following recent MRs:
This completely breaks CI on master
, so this seems pretty serious.