linalg_test failing initially but passing with rerun; ferror.h is not placed correctly

[jchristopherson]

linalg_test failing initially but passing with rerun; ferror.h is not placed correctly

[jchristopherson]

Hi @barracuda156, interesting. I'll try and replicate the failure on my end. My first suspicion is related to how I use a lot of random number generation for creating test matrix cases and have had issues with those causing failures in the past. The issues I've run into relate to the random generation producing matrices sufficiently close to singular to cause problems in routines that don't play well with such matrices. I'll let you know what I find.

As for the ferror header issue, I'm actually working on an update to that library now and will fix as part of that update. Stay tuned, but this shouldn't be too long before I get it fixed.

[jchristopherson]

Hi @barracuda156. I've made updates that should fix the issues you've highlighted. I have some other open action items, but feel free to test things out and let me know if you still have any issues.

[jchristopherson]

There are a few issues with the complex-valued routines. However, I added the BUILD_C_API flag as an option to allow the user to choose whether they wanted it or not regardless of the status of any of the routines.

[jchristopherson]

OK. Thanks for letting me know. I'll dig into this as well.

[jchristopherson]

That was an easy fix. It was related to the underlying ferror library having a change. The error resulted from the install script of that library. I'd recommend starting with a clean build and you should be fine.

[jchristopherson]

Good to know. I set up a Linux and MacOS CI process using both gfortran and ifort. I’ll keep an eye on this to help monitor for issues.

Thanks for your help and input so far!

[barracuda156]

@jchristopherson Thank you for responding!

(I intended to bring linalg into Macports, but looks like I rather wait until your update first.)

[barracuda156]

Hi @barracuda156. I've made updates that should fix the issues you've highlighted. I have some other open action items, but feel free to test things out and let me know if you still have any issues.

Thank you. By the way, is there any problem with BUILD_C_API at the moment or you just added an option to build it or not?

[barracuda156]

@jchristopherson So the build error due to ferror.h is fixed (expectedly). Tests still fail and pass randomly:

10:linalg-61747861b4491d21d637debb9c7fb86fced738c4 svacchanda$ /usr/bin/sudo ctest test --rerun-failed --output-on-failure
Test project /opt/local/var/macports/build/_opt_PPCRosettaPorts_math_linalg/linalg/work/jchristopherson-linalg-6174786
    Start 1: linalg_test
1/1 Test #1: linalg_test ......................***Failed    3.25 sec
Test Failed: Asymmetric Eigen Values Test 2
STOP 1


0% tests passed, 1 tests failed out of 1

Total Test time (real) =   3.28 sec

The following tests FAILED:
	  1 - linalg_test (Failed)
Errors while running CTest
10:linalg-61747861b4491d21d637debb9c7fb86fced738c4 svacchanda$ /usr/bin/sudo ctest test --rerun-failed --output-on-failure
Test project /opt/local/var/macports/build/_opt_PPCRosettaPorts_math_linalg/linalg/work/jchristopherson-linalg-6174786
    Start 1: linalg_test
1/1 Test #1: linalg_test ......................   Passed    3.26 sec

100% tests passed, 0 tests failed out of 1

Total Test time (real) =   3.29 sec

These two are successive runs from the same build.

[barracuda156]

Also install fails:

CMake Error at _deps/ferror-build/install/cmake_install.cmake:81 (file):
  file INSTALL cannot find
  "/opt/local/var/macports/build/_opt_PPCRosettaPorts_math_linalg/linalg/work/linalg-61747861b4491d21d637debb9c7fb86fced738c4/_deps/ferror-src/doc/html":
  No such file or directory.
Call Stack (most recent call first):
  _deps/ferror-build/cmake_install.cmake:52 (include)
  dependencies/ferror/cmake_install.cmake:42 (include)
  dependencies/cmake_install.cmake:52 (include)
  cmake_install.cmake:47 (include)


make: *** [install/fast] Error 1

[barracuda156]

@jchristopherson Tests issue might be OpenBLAS-related. I have set the build to use vecLibFort instead, and all looks good now from the first run:

--->  Testing linalg
Executing:  cd "/opt/local/var/macports/build/_opt_PPCRosettaPorts_math_linalg/linalg/work/build" && ctest test 
Test project /opt/local/var/macports/build/_opt_PPCRosettaPorts_math_linalg/linalg/work/build
    Start 1: linalg_test
1/3 Test #1: linalg_test ......................   Passed    0.75 sec
    Start 2: linalg_c_test
2/3 Test #2: linalg_c_test ....................   Passed    0.28 sec
    Start 3: fth_test
3/3 Test #3: fth_test .........................   Passed    0.02 sec

100% tests passed, 0 tests failed out of 3

Total Test time (real) =   1.13 sec

[jchristopherson]

Hi @barracuda156. I've finished my updates to the library, including adding additional test cases and CI processes. If everything is working properly for you, I'm inclined to close out this issue. I'll wait for your response, however.

[barracuda156]

Hi @barracuda156. I've finished my updates to the library, including adding additional test cases and CI processes. If everything is working properly for you, I'm inclined to close out this issue. I'll wait for your response, however.

@jchristopherson Let me try today. Thank you.

[barracuda156]

Thanks, all is good, closing.