On Fri, May 18, 2001 at 11:43:01AM -0500, John R. Jackson wrote:
> >I was never able to compile GNU tar with gcc 2.8.1 on Solaris.
> >But I work fine if compiled with SUN cc.
>
> I retried my test case using GNU tar 1.13.19 built with Sun cc and
> with gcc-2.95.3. Both work fine. So it was a bug with 2.8.1.
I don't know what the official word is on gcc, but I can report
my own results.
I've been building and using gcc in various native and cross
configurations since early gcc-2.3 (built on a 386-16 with
4M of RAM, it took all night to build) and the versions I've
had the best luck with are gcc-2.7.2.1 through 2.7.2.3, and
2.95.3. 2.95.2 was very good, too, but it won't handle some
of the recent c++ code I need it to build.
Some of the egcs versions worked pretty well for me, too, but
egcs is officially deprecated, as I understand it, in favor of
gcc-2.95.
> As Dan Wilder said, the important point is that you can never do
> enough backup/restore testing, no matter what [archiving utility]
> you use.
Thanks, John, for emphasizing that. I believe the point is too
often neglected.
A correlary: not testing at all is never sufficient. Even a little
bit of testing is a whole lot better than none, but there are gains to
be had from even more testing.
One thing I always do in my amanda run scripts, is give
"amverify" a crack at the tape, with the result emailed to me.
The extra wear and tear on tape and tape drive seems more than
justified by the confirmation that the darned thing was read
back successfully, at least once, for every backup.
--
Dan Wilder
