r184308 - [Driver] Don't forward source file input args to gcc

[Reid Kleckner]

On Wed, Jun 19, 2013 at 1:19 PM, Charles Davis <cdavis5x at gmail.com> wrote:

>
> On Jun 19, 2013, at 9:09 AM, Reid Kleckner wrote:
>
> > Author: rnk
> > Date: Wed Jun 19 10:09:06 2013
> > New Revision: 184308
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=184308&view=rev
> > Log:
> > [Driver] Don't forward source file input args to gcc
> >
> > gcc's inputs are already added by the InputInfoList passed to
> > Action::ConstructJob.
> >
> > Fixes a regression from r183989.  This was manifesting when targetting
> > mingw as an extra input argument to gcc when assembling.  It presumably
> > affects other situations where clang calls gcc.
> >
> > Prior to r183989, forwardToGCC() was returning false because the INPUT
> > option defined in OptParser.td had the DriverOption flag set on it.
> > LLVM's Option library does not set this flag for INPUT.
> >
> > Reviewers: espindola
> >
> > Differential Revision: http://llvm-reviews.chandlerc.com/D999
> >
> > Modified:
> >    cfe/trunk/lib/Driver/Tools.cpp
> >    cfe/trunk/test/Driver/no-integrated-as-win.c
> >
> > Modified: cfe/trunk/lib/Driver/Tools.cpp
> > URL:
> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Driver/Tools.cpp?rev=184308&r1=184307&r2=184308&view=diff
> >
> ==============================================================================
> > --- cfe/trunk/lib/Driver/Tools.cpp (original)
> > +++ cfe/trunk/lib/Driver/Tools.cpp Wed Jun 19 10:09:06 2013
> > @@ -225,7 +225,10 @@ static void addProfileRT(const ToolChain
> > }
> >
> > static bool forwardToGCC(const Option &O) {
> > -  return !O.hasFlag(options::NoForward) &&
> > +  // Don't forward inputs from the original command line.  They are
> added from
> > +  // InputInfoList.
> > +  return !O.getKind() == Option::InputClass &&
> Did you mean:
>
>   return O.getKind() != Option::InputClass &&
>                      ^
>
> ? Am I the only one who noticed this?


No, someone else mailed me about this off list.  If someone can fix it
quick, please do, it's not convenient for me right now.


> How are the tests even passing with this?


There's only one driver test that exercises this, which I just added.  It's
loose because I was trying to solve a problem of too many args, not too
few.  I can beef it up tomorrow.


> And why didn't Clang warn on this for you? (Want me to file a bug?)


As Matt said, we have the warning, but it must not be on.  I don't have
clang self-hosting on Windows, but a bot could've told me about it.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130619/8f2ecceb/attachment.html>