From: Jeffrey A Law <law@hurl.cygnus.com>

  Marc Espie writes:
  > I haven't really been able to follow thru the code (and my current build
  > is not finished yet, so I can't do the check myself), but your recent patch
  >
  > to handle weak correctly does look like __attribute__((weak)) will work
  > correctly only if HANDLE_PRAGMA_WEAK is asserted.
  >
  > This is a problem, as HANDLE_PRAGMA_WEAK is a choice, not a feature (at
  > least, that's what the gcc.info documentation seems to imply for
  > HANDLE_SYSV_PRAGMA).
When HANDLE_PRAGMA_WEAK is false, the code will continue to fail as it has
failed for years.  Not perfect by any means.  Feel free to fix it.  It is
still a choice, selecting that choice is the only way to get functional
weak support right now.

What needs to happen long term is the generic weak/alias code needs to become
unconditional and only the pragma parsing of that code be conditional.

1 files changed, 5 insertions, 0 deletions

diff --git a/gnu/usr.bin/gcc/config/openbsd.h b/gnu/usr.bin/gcc/config/openbsd.h
index af8358eb153..d865c8c03f1 100644
--- a/gnu/usr.bin/gcc/config/openbsd.h
+++ b/gnu/usr.bin/gcc/config/openbsd.h
@@ -300,3 +300,8 @@ do {									 \
    #pragma weak.  */
 #define GTHREAD_USE_WEAK 0
 
+/* bug work around: we don't want to support #pragma weak, but the current
+   code layout needs HANDLE_PRAGMA_WEAK asserted for __attribute((weak)) to
+   work.  On the other hand, we don't define HANDLE_PRAGMA_WEAK directly,
+   as this depends on a few other details as well...  */
+#define HANDLE_SYSV_PRAGMA