OpenBSD: testmod_mathcheck fails

[petermax2]

 19/104 Test  #19: testmod_mathcheck ..................***Failed    0.01 sec
elektraFtoA: val: 1.5, ret: 1.5
elektraEFtoF: string: 1.5, ret: 1.5
MATHCHECK      TESTS
==================

/root/Projekte/libelektra/src/plugins/mathcheck/testmod_mathcheck.c:73: error in main: error
/root/Projekte/libelektra/src/plugins/mathcheck/testmod_mathcheck.c:77: error in main: error
/root/Projekte/libelektra/src/plugins/mathcheck/testmod_mathcheck.c:81: error in main: error
/root/Projekte/libelektra/src/plugins/mathcheck/testmod_mathcheck.c:84: error in main: error
/root/Projekte/libelektra/src/plugins/mathcheck/testmod_mathcheck.c:87: error in main: error

testmod_mathcheck RESULTS: 60 test(s) done. 5 error(s).

[markus2330]

@tom-wa Nr. 3, so its definitely not an icc bug.

[markus2330]

@tom-wa should we exclude it for 0.8.16? (tag as experimental?)

[markus2330]

Also happened again on debian: http://community.markus-raab.org:8080/job/elektra-multiconfig-gcc-unstable/build_type=Debug,gcc_version=5,plugins=ALL/68/console

There are also

Memory checking results:
Uninitialized Memory Conditional - 79
Uninitialized Memory Read - 71

which is a clear indiciation of the underlying issues. Did you try to run valgrind on gcc5 debug?

[tom-wa]

yeah, please tag it as experimental, the plugin will get some rework anyway together with spec and conditionals for the next release (new syntax for relative and absolute keys + some cleanup and refactoring because it became quite messy after adding some of the newer features)

[tom-wa]

just ran it with valgrind gcc5 and debug build, no issues or memleaks. can you try to reproduce some of the issues locally on your boxes ?

[markus2330]

I think openbsd is the easiest way to reproduce it. On debian unstable it is non-deterministic.

[petermax2]

It also happens on Fedora.

MATHCHECK      TESTS
==================

src/plugins/mathcheck/testmod_mathcheck.c:73: error in main: error
src/plugins/mathcheck/testmod_mathcheck.c:77: error in main: error
src/plugins/mathcheck/testmod_mathcheck.c:81: error in main: error
src/plugins/mathcheck/testmod_mathcheck.c:84: error in main: error
src/plugins/mathcheck/testmod_mathcheck.c:87: error in main: error

testmod_mathcheck RESULTS: 60 test(s) done. 5 error(s).
elektraFtoA: val: 1,5, ret: 1.5
elektraEFtoF: string: 1.5, ret: 1,5

[markus2330]

Thanks for reporting!

[markus2330]

Will be solved later with other improvements in mathcheck.

[markus2330]

As discussed, check if its works with compilation variants for other data types (float, int, unsigned int, maybe even fixed point). If it works, its a floating point related problem.