[cfe-commits] r148138 - in /cfe/trunk: include/clang/Driver/ lib/Basic/ lib/Driver/ test/ARCMT/ test/CodeGen/ test/CodeGenCXX/ test/Coverage/ test/Driver/ test/Frontend/ test/Headers/ test/Index/ test/Misc/ test/PCH/ test/Preprocessor/ tools/driv

[Chandler Carruth]

On Fri, Jan 20, 2012 at 2:04 PM, Sebastian Pop <spop at codeaurora.org> wrote:

> On Fri, Jan 13, 2012 at 5:20 PM, Eli Friedman <eli.friedman at gmail.com>
> wrote:
> > On Fri, Jan 13, 2012 at 2:05 PM, Sebastian Pop <spop at codeaurora.org>
> wrote:
> >> Hi Eli,
> >>
> >> On Fri, Jan 13, 2012 at 3:38 PM, Eli Friedman <eli.friedman at gmail.com>
> wrote:
> >>> I had to revert this in r148141; it was causing regression test
> >>> failures on Mac buildbots.  A transcript of the failures follows:
> >>>
> >>
> >> Thanks for reporting these fails.  Are these the only fails on the
> >> buildbot?
> >
> > Yes, just those two.
> >
> >> I would need some help on understanding what is a normal output and
> >> what is erroneous in this log file: i.e., what went wrong.
> >
> > the noexecstack.c test is crashing.  I haven't looked closely at the
> other test.
>
> I see that -triple flag is parsed somewhere in the clang front-end, so
> I removed the -triple alias that I added:
> +def triple : Flag<"-triple">, Alias<target>;
>
> Here is the wrong change in bindings.c:
> -// RUN: %clang -ccc-host-triple i386-unknown-unknown
> -ccc-print-bindings -ccc-no-clang-cpp -fsyntax-only -no-integrated-cpp
> %s 2> %t
> +// RUN:target -target i386-unknown-unknown -ccc-print-bindings
> -ccc-no-clang-cpp -fsyntax-only -no-integrated-cpp %s 2> %t
>
> It seems like I replaced %clang with target, so this test wouldn't pass.
> I am committing the amended patch.
>

It remains really unclear why this is sufficient as a rename. I believe
there are parts of Clang that are using this flag to mean the *host*, for
real. I think renaming it blindly to 'target' doesn't actually make it
represent the concept of a target. Have you audited the uses?

Also, this patch was never actually LG'ed on the list. It should have been
review sooner, but a ping would seem more appropriate than just
committing...
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20120120/0e0e2276/attachment.html>