#2818: failing tests with gcc-14
| Reporter: | Roland Haas |
| Status: | new |
| Milestone: | |
| Version: | |
| Type: | bug |
| Priority: | major |
| Component: | EinsteinToolkit thorn |
I see failing tests, at least TestProlongate maybe others, but not all of them, when compiling the ET (Landau or development) with gcc-14.
This
./utils/Scripts/MakeThornList --master thornlists/einsteintoolkit.th arrangements/CarpetX/TestProlongate/test/test_cc_co_o1.par >thornlists/test_cc_co_o1.th
produces a thornlist one can try this out with. Eg for test_natural_o7.diffs the test system reports
testprolongate-gf001_count.it000002.x.tsv: substantial differences
significant differences on 68 (out of 39) lines
missing 29 lines in new testprolongate-gf001_count.it000002.x.tsv
maximum absolute difference in column 2 is 0.0166666666666667
maximum relative difference in column 2 is 0.25
and looking at the log files there seems to be an issue that only a single refinement level is created compared to 2 when compiled with gcc-13:
INFO (CarpetX): Regridding...
INFO (CarpetX): Setting max_grid_size values for all levels before regridding
INFO (CarpetX): ErrorEst patch 0 level 0 INFO (CarpetX): ErrorEst patch 0 level 0 done. Set/clear/total=0/27000/27000=0%/100%/100% INFO (CarpetX): old levels 1, new levels 1
INFO (CarpetX): level 0: 1 boxes, 27000 cells (100%)
so this does not seem to be just some sort of roundoff error thing to me.
Diff of the interesting bit of the log file
INFO (CarpetX): Regridding...
INFO (CarpetX): Setting max_grid_size values for all levels before regridding
INFO (CarpetX): ErrorEst patch 0 level 0
-INFO (CarpetX): Erro
--
Ticket URL: https://bitbucket.org/einsteintoolkit/tickets/issues/2818/failing-tests-with-gcc-14