test_tic_toc unstable

[slayoo]

See e.g. (on a genuinely unrelated change) https://travis-ci.org/gnudatalanguage/gdl/jobs/368263543

[alaingdl]

I saw nothing related to "test_tic_toc" in these Travis' logs ! (but maybe it is too late ?)

[slayoo]

let's forget about this one and try anew

[opoplawski]

This seems to be very unstable in Fedora builds, e.g.:

        Start 198: test_tic_toc.pro
198: Test command: /builddir/build/BUILD/gdl-1.1.1-build/gdl-v1.1.1/build/src/gdl "-quiet" "-e" "if execute('test_tic_toc') ne 1 then exit, status=1"
198: Working Directory: /builddir/build/BUILD/gdl-1.1.1-build/gdl-v1.1.1/build/testsuite
198: Environment variables: 
198:  LC_COLLATE=C
198:  GDL_PATH=/builddir/build/BUILD/gdl-1.1.1-build/gdl-v1.1.1/testsuite/:/builddir/build/BUILD/gdl-1.1.1-build/gdl-v1.1.1/src/pro/
198:  GDL_STARTUP=
198:  IDL_STARTUP=
198: Test timeout computed to be: 3600
198: % Compiled module: TEST_TIC_TOC.
198: % Time elapsed : 9.0122223e-05 seconds.
198: Time elapsed in loop  0 :       0.16580081 seconds.
198: Time elapsed in loop  1 :       0.15046883 seconds.
198: Time elapsed in loop  2 :       0.18779302 seconds.
198: Time elapsed in loop  3 :       0.14626908 seconds.
198: Cumulative Time elapsed :       0.65033174 seconds.
198: Total Time elapsed      :       0.65163088 seconds.
198: Time Precision          :     0.0021707820.
198: % Compiled module: BANNER_FOR_TESTSUITE.
198: % Compiled module: GDL_IDL_FL.
198: % TEST_TIC_TOC: ====================================================
198: % TEST_TIC_TOC: =                                                  =
198: % TEST_TIC_TOC: =  1 errors encountered during TEST_TIC_TOC tests  =
198: % TEST_TIC_TOC: =                                                  =
198: % TEST_TIC_TOC: ====================================================
198/216 Test #198: test_tic_toc.pro ...................***Failed    1.17 sec

[GillesDuvert]

This test is IMHO nuts and should be removed. I'ma afraid that due to tic toc using GDL procedure, not C++ intrinsics, the reported time passed executing something depends on the kernel, the architecture of the machine and how the processor is loaded and a 10% variation of this time is by far not an error. It should have occured us that, as soon as the tic toc procedure was rewritten to produce real systimes, not bogus ones as before, its accuracy will show the small performance changes induced by the other running processes on the computer.