gregor herrmann: > On Sun, 03 Jan 2021 11:53:22 +0100, Niels Thykier wrote: > >> AFAICT, upstream ought to have exit'ed with a non-zero exit code here >> because the configure job have in fact failed (albeit deliberate) to >> prepare the next stage of the build. > […] >> From my PoV, I see this as bug in the upstream package and unactionable >> feature request in debhelper. > > TTBOMK, that's considered "best practice" in the CPAN universe, in > order to let the process continue (where "process"is installing > multiple distributions or especially CPAN smoke testers). > > Random examples from my SSD: > > libnet-ssleay-perl/Makefile.PL: exit 0; # according > http://wiki.cpantesters.org/wiki/CPANAuthorNotes this is best-practice when > "missing library" > > libterm-readline-gnu-perl/Makefile.PL:# exit 0 before creating the Makefile > to be CPAN Testers friendly > > libtext-qrcode-perl/Makefile.PL: exit 0; # tell cpan testers that this > is not a failure > > libxml-libxml-perl/Makefile.PL: exit 0; # 0 recommended by > http://cpantest.grango.org (Notes for CPAN Authors) > > > Cheers, > gregor >
Hi gregor, That was an "interesting" practice. I cannot say I am happy with it, but if it is "best practise" then I am willing to reconsider the "wontfix" if you (as in: "anyone") can provide a reliable method for detecting this "false success" that is specific to Makefile.PL (or the perl related build systems). That said, such a solution will be "globally" enabled (possibly in a new compat level) and accordingly it most be a general solution. ~Niels
