Re: fixinc for Solaris' pthread.h

[DJ Delorie <dj at redhat dot com>]

> It is Sun's feeble attempt to accommodate GCC.  When GCC
> runs, it always selects the union and the initializer is
> configured for the int64_t's.  So, either the initializers
> have to be adjusted to use `{0}' in place of `0' in the
> correct places, or these pad64_t typedefs have to get fixed.
> This entails fixing sys/types.h.  I think the complete fix
> is a big nuisance and that is why it doesn't get done.

Maybe I'm missing something, then.  For solaris 7-9, we already have a
fix that changes the initializer.  I'm only suggesting an *additional*
fix for 2.6, not to fix the padding, but to fix the *other* part of
the initializer.  In the files you sent, 2.6 has this:

// #define	PTHREAD6_MUTEX_INITIALIZER	{0, 0, 0}
#define	PTHREAD6_MUTEX_INITIALIZER	{{{0}}, {{{0}}},  PTHREAD_PADDING_INIT}

But in the 2.6 I'm looking at, I see this:

#define	PTHREAD6_MUTEX_INITIALIZER	{0, 0, 0}

This case is what I'm trying to accomodate.