Test data_InterRegFlow fails on ppc64el built with -O3 optimization level

[LocutusOfBorg]

Hello, as said in the title, the data_InterRegFlow test fails on ppc64el

See e.g. https://launchpadlibrarian.net/680095999/buildlog_ubuntu-mantic-ppc64el.opm-common_2023.04+ds-2_BUILDING.txt.gz

Note, this happens only on ppc64el with -O3 and works perfectly with -O2
See e.g. https://launchpad.net/ubuntu/+source/opm-common/2023.04+ds-2ubuntu1/+build/26477122/+files/buildlog_ubuntu-mantic-ppc64el.opm-common_2023.04+ds-2ubuntu1_BUILDING.txt.gz

146: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_InteHEAD
146: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
146: Test timeout computed to be: 1500
146: Running 21 test cases...
145: 
145: �[1;32;49m*** No errors detected
145: �[0;39;49m
143/177 Test #145: DoubHEAD .........................   Passed    0.36 sec
test 147
        Start 147: data_InterRegFlow

147: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_data_InterRegFlow
147: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
147: Test timeout computed to be: 1500
147: Running 5 test cases...
147: ./tests/test_data_InterRegFlow.cpp(275): �[1;31;49merror: in "InterReg_View/Copy_Move_Assign_Add": Moved-from source range must be invalid�[0;39;49m
147: 
147: �[1;31;49m*** 1 failure is detected in the test module "data_InterRegFlow"
147: �[0;39;49m
144/177 Test #147: data_InterRegFlow ................***Failed    0.01 sec
Running 5 test cases...
./tests/test_data_InterRegFlow.cpp(275): �[1;31;49merror: in "InterReg_View/Copy_Move_Assign_Add": Moved-from source range must be invalid�[0;39;49m

�[1;31;49m*** 1 failure is detected in the test module "data_InterRegFlow"
�[0;39;49m

test 148
        Start 148: data_InterRegFlowMap

148: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_data_InterRegFlowMap
148: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
148: Test timeout computed to be: 1500
148: Running 8 test cases...
148: 
148: �[1;32;49m*** No errors detected
148: �[0;39;49m
145/177 Test #148: data_InterRegFlowMap .............   Passed    0.01 sec
test 149
        Start 149: LinearisedOutputTable

149: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_LinearisedOutputTable
149: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
149: Test timeout computed to be: 1500
149: Running 8 test cases...
149: 
149: �[1;32;49m*** No errors detected
149: �[0;39;49m
146/177 Test #149: LinearisedOutputTable ............   Passed    0.01 sec
test 150
        Start 150: LogiHEAD

150: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_LogiHEAD
150: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
150: Test timeout computed to be: 1500
150: Running 3 test cases...
150: 
150: �[1;32;49m*** No errors detected
150: �[0;39;49m
147/177 Test #150: LogiHEAD .........................   Passed    0.01 sec
test 151
        Start 151: LGOData

151: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_LGOData
151: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
151: Test timeout computed to be: 1500
144: 
144: �[1;32;49m*** No errors detected
144: �[0;39;49m
148/177 Test #144: EclipseIO ........................   Passed    0.53 sec
test 152
        Start 152: OutputStream

152: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_OutputStream
152: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
152: Test timeout computed to be: 1500
152: Running 11 test cases...
152: 
152: �[1;32;49m*** No errors detected
152: �[0;39;49m
149/177 Test #152: OutputStream .....................   Passed    0.02 sec
test 153
        Start 153: regionCache

153: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_regionCache
153: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
153: Test timeout computed to be: 1500
151: Running 1 test case...
146: 
146: �[1;32;49m*** No errors detected
146: �[0;39;49m
150/177 Test #146: InteHEAD .........................   Passed    0.33 sec
test 154
        Start 154: PaddedOutputString

154: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_PaddedOutputString
154: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
154: Test timeout computed to be: 1500
153: Running 1 test case...
153: 
153: �[1;32;49m*** No errors detected
153: �[0;39;49m
151/177 Test #153: regionCache ......................   Passed    0.17 sec
test 155
        Start 155: Restart

155: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_Restart
155: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
155: Test timeout computed to be: 1500
113: 
113: �[1;32;49m*** No errors detected
113: �[0;39;49m
152/177 Test #113: SummaryConfigTests ...............   Passed    5.10 sec
test 156
        Start 156: RFT

156: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_RFT
156: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
156: Test timeout computed to be: 1500
154: Running 3 test cases...
154: 
154: �[1;32;49m*** No errors detected
154: �[0;39;49m
153/177 Test #154: PaddedOutputString ...............   Passed    0.01 sec
test 157
        Start 157: rst_netbalan

157: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_rst_netbalan
157: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
157: Test timeout computed to be: 1500
157: Running 8 test cases...
157: 
157: �[1;32;49m*** No errors detected
157: �[0;39;49m
154/177 Test #157: rst_netbalan .....................   Passed    0.01 sec
test 158
        Start 158: Solution

158: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_Solution
158: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
158: Test timeout computed to be: 1500
158: Running 3 test cases...
158: 
158: �[1;32;49m*** No errors detected
158: �[0;39;49m
155/177 Test #158: Solution .........................   Passed    0.01 sec
test 159
        Start 159: Inplace

159: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_Inplace
159: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
159: Test timeout computed to be: 1500
151: 
151: �[1;32;49m*** No errors detected
151: �[0;39;49m
156/177 Test #151: LGOData ..........................   Passed    0.23 sec
test 160
        Start 160: Summary

160: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_Summary
160: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
160: Test timeout computed to be: 1500
159: Running 1 test case...
159: 
159: �[1;32;49m*** No errors detected
159: �[0;39;49m
157/177 Test #159: Inplace ..........................   Passed    0.01 sec
test 161
        Start 161: Summary_Group

161: Test command: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/bin/test_Summary_Group
161: Working Directory: /<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu/tests
161: Test timeout computed to be: 1500
155: Running 12 test cases...
156: Running 16 test cases...
161: Running 1 test case...
161: 
161: �[1;32;49m*** No errors detected
161: �[0;39;49m
158/177 Test #161: Summary_Group ....................   Passed    0.09 sec
test 162
        Start 162: Tables
[...]
99% tests passed, 1 tests failed out of 177

Total Test time (real) =  25.82 sec

The following tests FAILED:
	147 - data_InterRegFlow (Failed)
Errors while running CTest
make[1]: *** [Makefile:74: test] Error 8
make[1]: Leaving directory '/<<PKGBUILDDIR>>/obj-powerpc64le-linux-gnu'
dh_auto_test: error: cd obj-powerpc64le-linux-gnu && make -j4 test ARGS\+=--verbose ARGS\+=-j4 returned exit code 2
make: *** [debian/rules:17: binary-arch] Error 25
dpkg-buildpackage: error: debian/rules binary-arch subprocess returned exit status 2
--------------------------------------------------------------------------------
Build finished at 2023-08-02T20:37:26Z

[bska]

Thank you for the report. I will look into it.

[LocutusOfBorg]

thanks!

[blattms]

I could not see any sign why the move constructor should make the object invalid in all situation. We only make sure that
elements.begin==elements.end holds but maybe do not make sure that InterRegFlow::bufferSize does return something that is not zero. If is zero, then the object seems to be valid and the test will fail.

[bska]

If [bufferSize()] is zero, then the object seems to be valid and the test will fail.

Sure, but if that happens it's a compiler error. bufferSize() returns 2*NumComponents and NumComponents is 5 by construction.

[blattms]

You are right in this regard. of course. I failed to see that.

[LocutusOfBorg]

Maybe we can reduce the test case and ask gcc people to fix?

[bska]

Maybe we can reduce the test case and ask gcc people to fix?

I'm not convinced that this actually is a compiler bug, but I am trying to reduce this to a small example that would be usable as a "repro". At minimum we can use that in the Compiler Explorer to investigate. Failing that, experts on Stack Overflow and elsewhere might bring additional insight.

[blattms]

If that were a compiler bug it should have happened to many code. Maybe some out of bounds writing.
I will try to get access to a matching porter box at Debian to take a deeper look once they have updated my GPG key.

[blattms]

To me the problem might not be in opm-common. I tested this on Debian's platti porter machine. If I add a simple unrelated print statement (std::cout<<"default<<std::endl;) before the return in function isValid, then all tests pass. So I guess it might be a problem of the gcc optimizer on ppc64el.