Bruno Haible <br...@clisp.org> writes: > Hi Marc, > > Please keep the mailing list in CC. > >> > Does it still do so if you remove the 'valgrind-tests' modules from the >> > list? >> > >> >> It does not. > > OK, then the problem is that gnulib-tool does not know that 'valgrind-tests' > is special. Fixed like this. ... > +def _isTestsModuleName(name: str) -> bool: > + '''Determine whether a module is a tests module, given its name.''' > + return name.endswith('-tests') and name != 'valgrind-tests'
Ouch. Shouldn't we rename 'valgrind-tests' instead? The naming was a clever hack long time ago, but this module always behaved bit weird (e.g., creating an empty gnulib tests/ sub-directory if a package doesn't use any gnulib test modules), and I sometime regret the decision. Having a pattern that *-tests are test modules seems reasonable to me. Whatever problem appears if we rename 'valgrind-tests' to, say, 'valgrindtests' we could fix. Am I missing something? Sometimes clever hacks aren't clever in the long run. /Simon
signature.asc
Description: PGP signature