Hi, I almost forgot that one, it still builds and works properly on powerpc :)
Charlène. On Sun, 22 Sep 2019 11:30:49 +0200 Charlene Wendling wrote: > Hi, > > There have been some out of list discussion about it. > > COMPILER_VERSION is unused in ports and we have PROPERTIES for the > same purpose. jca@ proposed to remove COMPILER_VERSION from MAKE_ENV > and test bulk the change. The issue being i can't do a (required) full > bulk myself. > > Meanwhile mkvtoolnix still doesn't build for a trivial reason, and if > no one is willing to test bulk the MAKE_ENV change, i think we should > at least unset COMPILER_VERSION for that port. > > Opinions? > > Charlène. > > On Wed, 21 Aug 2019 21:09:33 +0200 > Charlene Wendling wrote: > > > Hi, > > > > > http://build-failures.rhaalovely.net/sparc64/2019-08-18/multimedia/mkvtoolnix%2Cno_x11.log > > (no full macppc bulk report available for now) > > > > What happens behind the scene (--trace didn't help): > > > > - $WRKSRC/rake.d/config.rb loads the 'build-config' file, that says > > that "COMPILER_VERSION = 8.3.0" (fine) > > - $WRKSRC/Rakefile does various compiler version checks to determine > > proper build flags to use > > - Compiler versions are checked in $WRKSRC/rake.d/helpers.rb using > > Ruby's check_version() function, that seems to not like version > > numbers with letters at the start, as we can see in the log. > > The compiler version itself comes from the c() function... > > - ...that is defined in $WRKSRC/rake.d/config.rb, where the line 24 > > is interesting: > > > > var = (ENV[idx_s].nil? ? $config[idx.to_sym] : ENV[idx_s]).to_s > > > > It appears that this 'gcc4' is coming from our default MAKE_ENV: > > > > $ make show=MAKE_ENV > > [...] COMPILER_VERSION="gcc4" [...] > > (on amd64, it returns COMPILER_VERSION="clang") > > > > It did not break on amd64, because with clang, upstream uses the > > same compilation flags for all versions, so there is no version > > comparison done. > > > > As such, i've overridden bsd.port.mk's COMPILER_VERSION, letting > > upstream's build system doing its job. I decided to go for it for > > all archs, as this may prevent possible future breakage on > > base-clang archs. > > > > I did not bump REVISION because this version never built on > > ports-gcc archs and it causes no change on base-clang ones. > > > > This builds, and tests pass with flying colors, on amd64 and macppc > > [0]. > > > > Comments/feedback are welcome :) > > > > Charlène. > > > > > > [0] https://bin.charlenew.xyz/mkvtoolnix.log > > > > Index: Makefile =================================================================== RCS file: /cvs/ports/multimedia/mkvtoolnix/Makefile,v retrieving revision 1.93 diff -u -p -u -p -r1.93 Makefile --- Makefile 7 Nov 2019 12:57:57 -0000 1.93 +++ Makefile 8 Jan 2020 22:57:38 -0000 @@ -29,6 +29,11 @@ MODRUBY_RUNDEP= No # C++11 / C++14 COMPILER = base-clang ports-gcc +# Override bsd.port.mk's COMPILER_VERSION because 'gcc4' is incorrect +# with ports-gcc and not recognised as a version number by Ruby's +# check_version(), and is just 'clang' with base-clang. +COMPILER_VERSION= + BUILD_DEPENDS= devel/gettext,-tools \ devel/gtest \ textproc/docbook \
