Fix and simplify build of static libraries.

[asarhaddon]

gprconfig was generating no Archive_Builder attribute for
mips-linux-gnu, powerpc64le-linux-gnu and many other architectures,
because the regular expressions defining specific architectures and
the fallback were not covering all cases. Ensure that all cases are
covered by the fallback stanza.

Replace many handwritten copies with variable substitutions.

Merge i686-pc-mingw32 and x86_64-pc-mingw32 as already done for other
architectures with aliases.
Merge native x86 into the default stanza.

This merge request was
AdaCore/gprbuild#67 before the split to gprconfig-kb.

This merge request is causing, for at least the third time, a non-trivial merge conflict.

A recent conflict (d9e1487) introduces an error (unless "-ranlib" is a valid command). The change in this request would most probably have prevented such an error.

[Fabien-Chouteau]

HI @asarhaddon thanks for your contribution.

You should have a look at this change: d49ff15
It is similar to what your are doing here. Except for the use of $PREFIX which is, for some targets, different than, $TARGET.

[asarhaddon]

This commit should hopefully fix the initial problem (some Debian architectures are missing a static archiver).

A separate merge request (#4) fixes a probable typo.

You are right that the remaining idea is similar to d49ff15.

• It seems that the native Archiver stanza plays the role of catch-all stanza for unknown architectures. If so, it may be useful to copy each pattern from its separate Archiver stanza, in the same order, into the list of negated patterns for the native Archiver stanza. There already exists inconsistencies.
• Why not use "${PREFIX}ar" for both native and cross builds, at least on most architectures?
  • during a native build, "ar" and "${PREFIX}ar" are synonym.
  • during a cross build, "${PREFIX}ar" only is allowed.

[CLAassistant]

All committers have signed the CLA.

[t-14]

Unfortunately, due to a veritable zoo of architectures that AdaCore has to support, this patch is problematic for us as is. May I suggest that you just add specific triplets you would like to exclude, but leave the generic regexes intact?