Hmm, I'm not sure, I cargo-culted the le32 test and didn't think about the non-hermeticity, but this does seem like a bad issue that I should fix.
The test needs va_list, size_t and ptrdiff_t. Should I try to include clang's headers directly? They use include_next so they will pull in the default headers if I don't prevent them from doing so. I'd obviously fix the le32 test (and any other in this directory). What do you recommend? On Fri, Sep 12, 2014 at 5:37 PM, Nick Lewycky <[email protected]> wrote: > On 12 September 2014 15:53, Nick Lewycky <[email protected]> wrote: > >> On 12 September 2014 10:52, JF Bastien <[email protected]> wrote: >> >>> --- cfe/trunk/test/Driver/le64-unknown-unknown.cpp (added) >>> +++ cfe/trunk/test/Driver/le64-unknown-unknown.cpp Fri Sep 12 12:52:47 >>> 2014 >>> @@ -0,0 +1,137 @@ >>> +// RUN: %clang -target le64-unknown-nacl -### %s -emit-llvm-only -c >>> 2>&1 | FileCheck %s -check-prefix=ECHO >>> +// RUN: %clang -target le64-unknown-nacl %s -emit-llvm -S -c -o - | >>> FileCheck %s >>> + >>> +// ECHO: {{.*}} "-cc1" {{.*}}le64-unknown-unknown.c >>> + >>> +// Check platform defines >>> +#include <stdarg.h> >>> +#include <stddef.h> >>> >> >> This test is non-hermetic. Please test only the driver's behaviour in a >> driver test. >> > > ... or is it? We have our own stdarg.h and stddef.h in lib/Headers? Does > this use those? >
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
