checking changes for warning regressions

[Roland McGrath]

> The warnings have always been there for x32 build.  I investigated it
> while working on x32 and came to the same conclusion.   This is the
> known problem:

I wonder if __builtin_choose_expr behaves differently.
That would make some sense.

This reminds me of what I am pretty sure was a real warning regression
introduced in 2.16 that actually affects users.  

pcprofiledump.c:173:5: warning: format '%lx' expects argument of type 'long unsigned int', but argument 2 has type 'long long unsigned int' [-Wformat] 

This is PRIx64 vs return value of bswap_64.  bswap_64's return type must
match the type underlying uint64_t.  But you changed it from long to long
long unconditionally, while x86-64 (-m64) int64_t is still just long.


Thanks,
Roland