------- Comment #2 from ghazi at gcc dot gnu dot org 2005-11-29 21:46 -------
Hmm this is convoluted, but I think I know what's going on:
We're running the builtin fprintf check. I recently added a small sanity check
to ensure that fprintf_unlocked also works. Now we're getting an unresolved
symbol calling fputs_unlocked.
But to ensure that these _unlocked style calls don't result in unresolved
symbols I had only added cases that should have been completely optimized away.
E.g. fprintf_unlocked(stream, "") which should become nothing.
Now hpux defines DONT_HAVE_FPUTC_UNLOCKED, notice that's fputC_unlocked not
fputS_unlocked which is the unresolved symbol we get. But we have this code at
the top of fold_builtin_fputs:
/* If the return value is used, or the replacement _DECL isn't
initialized, don't do the transformation. */
if (!ignore || !fn_fputc || !fn_fwrite)
return 0;
So the solution is to split these checks and move the !fn_fputc || !fn_fwrite
check later on where we actually attempt to use them. That way the path which
eliminates zero-length strings will still be executed.
I'll put together a patch.
--
ghazi at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
AssignedTo|unassigned at gcc dot gnu |ghazi at gcc dot gnu dot org
|dot org |
Status|UNCONFIRMED |ASSIGNED
Ever Confirmed|0 |1
Last reconfirmed|0000-00-00 00:00:00 |2005-11-29 21:46:13
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25158
