[MacPorts] #65498: gcc10-bootstrap fails on Rosetta due to wrong archs in dependencies

[MacPorts]

#65498: gcc10-bootstrap fails on Rosetta due to wrong archs in dependencies
------------------------------+-------------------------------------------
  Reporter:  barracuda156     |      Owner:  catap
      Type:  defect           |     Status:  assigned
  Priority:  Normal           |  Milestone:
 Component:  ports            |    Version:  2.7.2
Resolution:                   |   Keywords:  powerpc, rosetta, snowleopard
      Port:  gcc10-bootstrap  |
------------------------------+-------------------------------------------

Comment (by catap):

 This issue seems as good place for notes.

 The root cause of this issue that it tries to build gmp as:
 {{{
   Host type:         none-apple-darwin10
   ABI:               standard
  }}}
 because `Makefile.in` is overwrite provided host and target to `none-bla-
 bla` like that:
 {{{
         $(SHELL) \
           $$s/$$module_srcdir/configure \
           --srcdir=$${topdir}/$$module_srcdir \
           $(HOST_CONFIGARGS) --build=${build_alias}
 --host=none-${host_vendor}-${host_os} \
           --target=none-${host_vendor}-${host_os} --disable-shared
 LEX="touch lex.yy.c" \
           || exit 1
 }}}

 anyway, I may make a patch that removes this that won't solve an issue
 because GMP requires host to be specified up to target processor, see:
 https://github.com/macports/macports-
 ports/blob/43fc13d3e82f99ed563c4564cc6946ed75e75674/devel/gmp/Portfile#L117-L121

 Thus, such setup brokes GCC :)

 And to make things more tricky, GMP resets all provided CFLAGS and
 CXXFLAGS on configure stage, and without -arch it fails with error
 {{{
 checking if the assembler needs r on registers... configure: error:
 neither "mtctr 6" nor "mtctr r6" works
 }}}
 because gcc-4.2 on 10.6 builds x86_64 by default, and not ppc :)

-- 
Ticket URL: <https://trac.macports.org/ticket/65498#comment:5>
MacPorts <https://www.macports.org/>
Ports system for macOS