Maniphest T65143

ctest: bmesh_boolean test fails on x86
Closed, Resolved

Assigned To
Brecht Van Lommel (brecht)
Authored By
Ray Molenkamp (LazyDodo)
May 26 2019, 10:04 AM
Tags
  • BF Blender
  • Tracker Curfew
Subscribers
Brecht Van Lommel (brecht)
Campbell Barton (campbellbarton)
Ray Molenkamp (LazyDodo)

Description

System Information
Operating system: Windows X86 + Linux X86
Graphics card: N/A

Broken: Latest Master 32 bit
Worked: Latest Master 64 bit

Short description of error

the bmesh_boolean test fails on both windows and linux on 32 bit builds.

Additional information

The reported memleak doesn't seem to trigger on the x64 builds.

Blender 2.80 (sub 72) (hash f18373a9ab1a built Sun 05/26/2019 12:09 AM)
found bundled python: D:\a\1\build_windows_Full_x86_vc15_Release\bin\Release\2.80\python
Read blend: D:/a/1/s/../lib/tests/modeling/bool_regression.blend
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Info: Deleted 1 object(s)
Error: Not freed memory blocks: 3, total unfreed memory 0.002487 MB
D:/a/1/s/../lib/tests/modeling/bool_regression.blend\run_tests:4: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses
Traceback (most recent call last):
  File "D:/a/1/s/../lib/tests/modeling/bool_regression.blend\run_tests", line 38, in <module>
SystemExit: 1
Run test: intersect_boolean(operation="UNION") on Cubecube selecting F 0 1 2 3 4 5
Success
Info: Deleted 1 object(s)
Run test: intersect_boolean(operation="INTERSECT") on Cubecube selecting F 0 1 2 3 4 5
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect_boolean(operation="DIFFERENCE") on Cubecube selecting F 0 1 2 3 4 5
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect(separate_mode="CUT") on Cubecube selecting F 0 1 2 3 4 5
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect(separate_mode="ALL") on Cubecube selecting F 0 1 2 3 4 5
Fail Edge Mismatch
Info: Deleted 1 object(s)
Run test: intersect(separate_mode="NONE") on Cubecube selecting F 0 1 2 3 4 5
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect(mode="SELECT" separate_mode="NONE") on Cubecube selecting F 0 1 2 3 4 5 6 7 8 9 10 11
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect_boolean(operation="UNION") on Cubecone selecting F 6 7 8 9 10
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Run test: intersect_boolean(operation="UNION") on Cubecones selecting F 0 1 2 3 4 5
Fail Loop Vert Mismatch In Poly Test
Info: Deleted 1 object(s)
Ran 9 tests, 8 failed
Tests Failed

Exact steps for others to reproduce the error
run the bmesh_boolean test or load bool_regression.blend manually and click run script.

Related Objects

Event Timeline

Ray Molenkamp (LazyDodo) created this task.May 26 2019, 10:04 AM
Ray Molenkamp (LazyDodo) updated the task description.May 26 2019, 10:09 AM
Ray Molenkamp (LazyDodo) assigned this task to Campbell Barton (campbellbarton).May 27 2019, 6:14 PM
Ray Molenkamp (LazyDodo) lowered the priority of this task from 90 to High.
Ray Molenkamp (LazyDodo) added a subscriber: Campbell Barton (campbellbarton).

@Campbell Barton (campbellbarton) mind taking a peek here?

Brecht Van Lommel (brecht) lowered the priority of this task from High to 50.May 27 2019, 7:38 PM
Brecht Van Lommel (brecht) added a subscriber: Brecht Van Lommel (brecht).

I don't think a memory leak on x86 is high priority as in must be fixed for the 2.80 release. Still good to fix to make the tests run well.

Ray Molenkamp (LazyDodo) added a comment.May 27 2019, 7:43 PM

The actual tests are failing, the memleak is probably just a symptom.

Ray Molenkamp (LazyDodo) added a comment.May 28 2019, 3:31 AM

I looked a little further into this, the tests are failing because the geometry generated is actually different between x86 and x64 (errors out during CD_MPOLY compare in customdata_compare) the result doesn't look wrong, up to you if you want to keep it at medium.

Given you thought it was just a 3kb mem leak, the medium prio seems rather generous.

Brecht Van Lommel (brecht) lowered the priority of this task from 50 to Low.May 29 2019, 1:06 PM

It's medium priority in the sense that it would be good to have all our tests passing, though the actual bug is not that important probably.

It might be an issue with extended precision for floats on 32 bits or something like that, which wouldn't really make it a bug.

I want to drop 32 bit support right after the 2.80 release though, so fixing this a harmless issue for a test is not that important.

Dalai Felinto (dfelinto) removed Campbell Barton (campbellbarton) as the assignee of this task.Dec 23 2019, 4:34 PM
Dalai Felinto (dfelinto) added a project: Tracker Curfew.
Brecht Van Lommel (brecht) closed this task as Resolved.Jan 27 2020, 11:50 AM
Brecht Van Lommel (brecht) claimed this task.

We dropped 32 bit support, so not going to spend time on this anymore.