https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113877
--- Comment #3 from simon at pushface dot org --- I’ve now found the related documentation[1]. I think the conclusion is that gnatchop is working-as-designed, and that this PR should be marked WONTFIX (or INVALID), as seems best. Sorry for the noise. [1] https://docs.adacore.com/live/wave/gnat_ugn/html/gnat_ugn/gnat_ugn/the_gnat_compilation_model.html#operating-gnatchop-in-compilation-mode
