IBM long double fixes

[Alan Modra]

On Tue, Jun 25, 2013 at 01:31:36PM +0000, Joseph S. Myers wrote:
> On Tue, 25 Jun 2013, Alan Modra wrote:
> 
> > 	* sysdeps/ieee754/ldbl-128ibm/e_rem_pio2l.c: Comment fix.
> 
> This comment fix can go in straight away as obvious.
> 
> Fixes for bugs that were user-visible in past releases should have an 
> associated bug in Bugzilla with [BZ #N] in the ChangeLog entry and the bug 
> number listed in NEWS as fixed (if the bug was fully fixed), and testcases 
> for the testsuite may be a good idea as well (though we're less strict 
> about that for system-specific bugs).  I know of at least bugs 5268 and 
> 14551 as known IBM long double bugs in glibc, although the former doesn't 
> give a self-contained testcase so it's not clear what the bug is supposed 
> to be there.  I don't know if those bugs correspond to some of what you're 
> fixing or what other bugs might need to be filed to cover the fixes.

There is some chance that 5268 is fixed by
http://sourceware.org/ml/libc-alpha/2013-06/msg00919.html
but it doesn't fix 14551.  I didn't develop the denormal fixes in
response to any particular bugzilla, but rather from simply browsing
the source when making changes for little-endian support, and noticing
nastiness.  In fact, these bug fixes are a distraction from my main
aim, which is little-endian support.

Do you really have to file a bugzilla to fix bugs around here?

-- 
Alan Modra
Australia Development Lab, IBM