On Tue, Dec 6, 2016 at 9:01 AM, Florian Weimer <fweimer@xxxxxxxxxx> wrote: > On 12/06/2016 08:27 AM, Ralf Corsepius wrote: >> >> On 12/06/2016 04:44 AM, Orion Poplawski wrote: >> >>> I've run into a reasonable number of failures of (custom) configure >>> checks due to this. Mainly calling exit() without including stdlib.h. >>> So keep alert everyone. Unfortunately this can lead to unpredictable >>> and perhaps hard to detect changes. >> >> >> It's worse. This change contradicts autoconf's working principles. > > > I doubt that it's deliberate that autoconf attempts to compile invalid C > code. > >> It >> causes configure-checks to produce bogus results and to produce >> mal-built packages. > > > These packages are unportable and will fail with other C compilers. > >> That said, I am vehemently opposed to this step and ask the persons in >> charge to revert it. > > > We could probably help with identifying packages affected by this change, > but after seventeen years, it's time to clean up these broken autoconf > checks. Totally agree here. Florian, why only %__global_cflags were changed and not %optflags? Many of packages which don't do %configure, use %optflags. > > Thanks, > Florian > > _______________________________________________ > devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx > To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx -- -Igor Gnatenko _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
