R devel - Apr 2003 - RH 9 and R 1.7.0 Beta - response from Red Hat

Attached is Red Hat's judgement on my bug report for the missing value
corruption problem that appears in RH 9.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=88174

Martyn
-------------- next part --------------
An embedded message was scrubbed...
From: bugzilla@redhat.com
Subject: [Bug 88174] Optimization causes corruption of NA value in the R
	language
Date: Mon, 7 Apr 2003 12:13:24 -0400
Size: 1998
Url:
https://www.stat.math.ethz.ch/pipermail/r-devel/attachments/20030407/b0300c56/attachment.txt

>-----Original Message-----
>From: Martyn Plummer [mailto:plummer@iarc.fr] 
>Sent: Monday, April 07, 2003 11:25 AM
>To: r-devel@r-project.org
>Cc: Marc Schwartz; Peter Dalgaard
>Subject: RH 9 and R 1.7.0 Beta - response from Red Hat
>
>
>Attached is Red Hat's judgement on my bug report for the 
>missing value corruption problem that appears in RH 9.
>
>https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=88174
>
>Martyn

Martyn,

First, my apologies.  In hindsight, I should have sent you an e-mail
on Friday to inform you that Peter had contacted me. I was not aware
at the time that you had RH 9 running yet and so had not considered
the possibility that you might be working on the bug in parallel.
Sorry.

Thanks for confirming the problem and for posting the code and the bug
to RH.  I was on the RH web site reading Jakub's reply to your post as
your message above came through. 

There are other bugs posted there attributed to real.c, with comments
to the effect that the rewrite of real.c for gcc 3.4 CVS was
undertaken specifically to fix some of them. Jakub's reply would
suggest that they went ahead and incorporated some or all of the 3.4
real.c updates into 3.3.

Given his comments, it is curious however that this behavior is not
seen elsewhere using 3.2.2, unless other distros have included the
3.3/3.4 real.c changes or some other workaround as an incremental bug
fix into 3.2.2. This would possibly fit with my prior comments
regarding distro specific gcc version issues, since Paul did not see
this on Mandrake.

No response though by Jakub to the -pedantic issue, which is a
curiosity.

Best regards,

Marc Schwartz