[cfe-commits] r149083 - in /cfe/trunk: include/clang/Driver/ToolChain.h lib/Driver/Driver.cpp lib/Driver/ToolChain.cpp lib/Driver/ToolChains.cpp lib/Driver/ToolChains.h lib/Driver/WindowsToolChain.cpp test/Driver/prefixed-tools.c test/Driver/pref

[Joerg Sonnenberger]

On Fri, Jan 27, 2012 at 03:16:17PM -0800, Douglas Gregor wrote:
> > Reverting just this change is inacceptable. It fixes a major regression
> > for cross-compiling support from your earlier changes, without
> > introducing a bag of new ones.
> 
> There is a legitimate concern that your change is not the right
> direction. Please revert, and we can discuss how best to implement
> this functionality.

That doesn't answer the question of how to deal with the original change
that broke it. So before randonly reverting changes -- where should the
user triple belong?

(a) Driver. Old location before the driver refactoring with FIXME as it
required parsing the arguments.

(b) Toolchain. That's what implemented in this change. This also allows
platforms to deal with more limited toolchains, e.g. old as doesn't
support --32 if built only for i386.

Just reverting results in a regression. The functionality itself is
pretty much independent of the general cleanup.

There is a somewhat general issue of how much the current distinction
between Driver and Toolchain data really makes sense. E.g. can you
actually sanely have more than one toolchain per driver without also
changing sysroot etc. Moving potential wards around doesn't answer that
question.

Joerg