Make CMakeMake respect the toolchainopts when selecting build type

[Micket]

See title

[Flamefire]

LGTM

[Micket]

Test report by @Micket

Overview of tested easyconfigs (in order)

• SUCCESS CGAL-5.6.1-GCCcore-13.3.0.eb
• SUCCESS fmt-9.1.0-GCCcore-12.2.0.eb
• SUCCESS graphite2-1.3.14-GCCcore-13.3.0.eb
• SUCCESS json-c-0.17-GCCcore-13.3.0.eb
• SUCCESS libgit2-1.8.1-GCCcore-13.3.0.eb
• SUCCESS libheif-1.17.6-GCC-12.3.0.eb
• SUCCESS nvtop-3.1.0-GCCcore-12.3.0.eb
• SUCCESS Qhull-2020.2-GCCcore-13.3.0.eb
• SUCCESS RAxML-NG-1.2.0-GCC-12.3.0.eb
• SUCCESS SentencePiece-0.2.0-GCC-13.2.0.eb
• SUCCESS snappy-1.1.9-GCCcore-12.2.0.eb
• SUCCESS SRA-Toolkit-3.1.1-gompi-2023b.eb
• SUCCESS x265-3.6-GCCcore-13.3.0.eb
• SUCCESS yaml-cpp-0.8.0-GCCcore-13.2.0.eb

Build succeeded for 14 out of 14 (14 easyconfigs in total)
alvis-skylake-build - Linux Rocky Linux 8.9, x86_64, Intel Xeon Processor (Skylake, IBRS, no TSX), Python 3.6.8
See https://gist.github.com/Micket/1c4a1ac142e565f3b90e7402d981a30d for a full test report.

[Micket]

Will repeat the tests above with debug: True

[Micket]

Test report by @Micket

Overview of tested easyconfigs (in order)

• SUCCESS CGAL-5.6.1-GCCcore-13.3.0.eb
• SUCCESS fmt-9.1.0-GCCcore-12.2.0.eb
• SUCCESS gmsh-4.12.2-foss-2023a.eb
• SUCCESS graphite2-1.3.14-GCCcore-13.3.0.eb
• SUCCESS json-c-0.17-GCCcore-13.3.0.eb
• SUCCESS libgit2-1.8.1-GCCcore-13.3.0.eb
• SUCCESS libheif-1.17.6-GCC-12.3.0.eb
• SUCCESS nvtop-3.1.0-GCCcore-12.3.0.eb
• SUCCESS Qhull-2020.2-GCCcore-13.3.0.eb
• SUCCESS RAxML-NG-1.2.0-GCC-12.3.0.eb
• SUCCESS SentencePiece-0.2.0-GCC-13.2.0.eb
• SUCCESS snappy-1.1.9-GCCcore-12.2.0.eb
• SUCCESS SRA-Toolkit-3.1.1-gompi-2023b.eb
• SUCCESS x265-3.6-GCCcore-13.3.0.eb
• SUCCESS yaml-cpp-0.8.0-GCCcore-13.2.0.eb

Build succeeded for 15 out of 15 (15 easyconfigs in total)
alvis-skylake-build - Linux Rocky Linux 8.9, x86_64, Intel Xeon Processor (Skylake, IBRS, no TSX), Python 3.6.8
See https://gist.github.com/Micket/52ae77ae0a426b7f7c6150762494784f for a full test report.

[Micket]

which also looked good to me, so now with noopt

[Micket]

Test report by @Micket

Overview of tested easyconfigs (in order)

• SUCCESS CGAL-5.6.1-GCCcore-13.3.0.eb
• FAIL (build issue) fmt-9.1.0-GCCcore-12.2.0.eb (partial log available at https://gist.github.com/Micket/be6063393b1e2f4c3a39eeae1ed66d4e)
• SUCCESS graphite2-1.3.14-GCCcore-13.3.0.eb
• SUCCESS json-c-0.17-GCCcore-13.3.0.eb
• SUCCESS libgit2-1.8.1-GCCcore-13.3.0.eb
• SUCCESS libheif-1.17.6-GCC-12.3.0.eb
• SUCCESS nvtop-3.1.0-GCCcore-12.3.0.eb
• FAIL (build issue) Qhull-2020.2-GCCcore-13.3.0.eb (partial log available at https://gist.github.com/Micket/81f9890a7a2a45883c5eed329140e555)
• SUCCESS RAxML-NG-1.2.0-GCC-12.3.0.eb
• SUCCESS SentencePiece-0.2.0-GCC-13.2.0.eb
• SUCCESS snappy-1.1.9-GCCcore-12.2.0.eb
• FAIL (build issue) SRA-Toolkit-3.1.1-gompi-2023b.eb (partial log available at https://gist.github.com/Micket/4219b353f70b8ddeda8b2ac6e53e3064)
• SUCCESS x265-3.6-GCCcore-13.3.0.eb
• FAIL (build issue) yaml-cpp-0.8.0-GCCcore-13.2.0.eb (partial log available at https://gist.github.com/Micket/57c56ca5378f1ec9705fe94ee04768a4)

Build succeeded for 10 out of 14 (14 easyconfigs in total)
alvis-skylake-build - Linux Rocky Linux 8.9, x86_64, Intel Xeon Processor (Skylake, IBRS, no TSX), Python 3.6.8
See https://gist.github.com/Micket/0cf19fb6e8f346eb94f0a500b100ed44 for a full test report.

[Micket]

also looked good, though the noopt flag is still setting --march=native which seems strange (though that is due to *FLAGS environment variables, not due to the cmake easyblock.

Failures above is just that the build scripts intentionally append a "d" at the end of the library names, except for SRA-Toolkit which the build just breaks with you run debug (really unclear as to what is happening there, but, not CMake's fault from what i can see in the logs).

I think this is good to go in.

[bartoldeman]

lgtm

[Flamefire]

also looked good, though the noopt flag is still setting --march=native which seems strange (though that is due to *FLAGS environment variables, not due to the cmake easyblock.

I guess EBs optarch is orthogonal to debug builds. And makes sense to me.

Failures above is just that the build scripts intentionally append a "d" at the end of the library names,

For those it might be better to use (potentially iterated) configureopts to build debug and release versions. It is intentional that the debug libs are named differently to allow them to coexist.

except for SRA-Toolkit which the build just breaks with you run debug (really unclear as to what is happening there, but, not CMake's fault from what i can see in the logs).

"Usage: vdb-dump [ ...] [options]" --> Looks like it messes up an invocation of a helper program. Probably a bug in their build scripts which we can report.