McXtrace: SOLEIL_CRISTAL and SOLEIL_MARS test values off after Fluorescence fix (#3088) #3091

The full McXtrace test suite on main (run 43027588188, triggered by #3089) fails with two values off:

The values are identical on Linux (gcc, classic and antlr), macOS (clang) and Windows (MSVC), over two attempts. Everything else in the suite passes (0 compile errors, 0 runtime errors).

Likely cause

Both instruments use FluoPowder, which does SHARE INHERIT Fluorescence. #3088 changed the cross-section normalisation and sampling in Fluorescence.comp and updated the test values of SOLEIL_ROCK and Test_FluoPowder, but not these two. Not verified by a before/after run of #3088.

Why #3088 was green

The "Check for modified instruments" step ran mxtest --instr=SOLEIL_ROCK,Test_FluoPowder --comp=Fluorescence, which selected only SOLEIL_ROCK, Test_FluoPowder and Test_Fluorescence. The --comp scope does not follow INHERIT, so instruments using FluoPowder were not run. #3089 touches the code generator, which triggers the full suite and exposed the mismatch.

To do

• Confirm that the new values are the physically intended result of the normalisation fix (a factor 2.5-3 change), then update the %Test lines.
• Consider letting the changed-component scope in mctest/mxtest follow INHERIT, so that users of derived components are tested too.

Unrelated noise in the same run: attempt 2 lost the two source-build jobs in "Setup MPI" to apt fetch failures from security.ubuntu.com.