Eric Blake wrote: > According to Jim Meyering on 8/15/2009 9:50 AM: >>> I'm not sure why some invocations work and others fail, but the point is >>> that update-copyright doesn't work as-is on cygwin 1.5; and it also has >>> the annoying failure path of either renaming or worse deleting the >>> original file. So I'm not sure how best to add skip logic to >>> test-update-copyright to avoid spurious test failures. >> >> Well, with that, adjusting the test script >> to detect and skip should be trivial. > > Thanks - that was enough to get going. Here's what I'm pushing (I don't > think you'll mind being listed in the ChangeLog). > >> However, eventually, (for improved portability) I want to convert >> update-copyright itself not to use that mechanism. I.e., to have the >> usual #!/usr/bin/perl first line, without any options. That's also in >> line with the goal of adding --help and --version options. > > Yes, it would be good to clean up the #! line. But we still have to > remember that #!/usr/bin/perl is not required to exist on end user's > machines. Coreutils even goes so far as to dynamically use $(PERL) during > the testsuite (check.mk's shell_or_perl_ function), exploiting the checks > for perl done during configure. But while packages like coreutils or > autoconf that already require perl during configure can get away with > that, I don't think it is worth bloating configure to add $(PERL) checking > to simpler packages like m4 that have no other perl dependency.
Agreed. > Subject: [PATCH] test-update-copyright: skip if perl is insufficient > > * tests/test-update-copyright.sh: Failure to run maintainer tool > should not cause testsuite failure on cygwin 1.5. Thanks for handling that, and for cleaning up the temporaries when skipping. With your added trap, at least the final rm $TMP* has been rendered redundant. Though technically, I suspect all of them may be removed, now.
