[compiler-rt] r204079 - PGO: Statically generate data structures

Duncan P. N. Exon Smith dexonsmith at apple.com
Mon Mar 17 16:54:22 PDT 2014


On Mar 17, 2014, at 4:43 PM, Chandler Carruth <chandlerc at google.com> wrote:

> 
> On Mon, Mar 17, 2014 at 4:40 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> On Mar 17, 2014, at 3:53 PM, Chandler Carruth <chandlerc at google.com> wrote:
> 
> >
> > On Mon, Mar 17, 2014 at 3:28 PM, Duncan P. N. Exon Smith <dexonsmith at apple.com> wrote:
> > I see
> >
> > #include <stdio.h>
> > #include <stdlib.h>
> >
> > at the top of the file.  Also, it compiles without warnings for me (and it works for me).
> >
> > Maybe fputc isn’t portable?  r204084 switches to putc.  Maybe it’s more portable?
> >
> > Why would it be more portable? This function was in C89, it has existed for approximately ever. The 'putc' function seems just a confusing and broken alternative. I see no good reason to ever use it...
> 
> 
> Reverted everything in r204088, because as Chandler suggested putc is strictly worse.
> 
> Also, tracked down the real problem:  we need have local headers for functions in compiler-rt, and there aren’t any stubs for fputc.  Maybe it worked  locally because I wasn’t bootstrapping?
> 
> Ah, that makes sense out of everything.
> 
> I'm not at all familiar with how the stub headers work... =/ Maybe someone else knows. Sorry I can't actually help with the fix here.

I’m hoping r204091 fixes this.



More information about the llvm-commits mailing list