Michael Stassen <[EMAIL PROTECTED]> writes:

> On Fri, 8 Mar 2002, Trond Eivind Glomsrød wrote:
> 
> > On Fri, 8 Mar 2002, Ireneusz Piasecki wrote:
> >
> > > Hi,
> > >
> > > It is very interesting, what are you wwriting, but:
> > >
> > >
> > > " To determine if you should be concerned about this compiler issue, execute
> > > gcc -v from the command prompt on your system. If the compiler reports
> > > version 2.96, then there is a problem (this is the case, for example on RH
> > > 7.x series or Mandrake 8.x). In this case, you should not try to compile
> > > your own binary before downgrading to one of the compilers mentioned above.
> > > You should also NOT use the MySQL server provided with your distribution --
> > > as this copy of MySQL was compiled with the same ill-advised compiler
> > > version. "
> > > this is from http://www.mysql.com/downloads/mysql-3.23.html
> > > I understud, don't compile mysql with 2.96 of gcc.
> > > I'm confiused.
> > >
> > > The query is: Compile server mysql with 2.96 or not ?
> >
> > 2.96RH. It's more widely used than gcc3 (which has had grave problems, but
> > have been stabilizing) and gcc 2.95 (which is also rather bad...)[1] would
> > definitely be the one I recommend. We trust it enough that when we upgrade
> > the compiler to gcc 3.1 someday in the future, the kernel will probably
> > stay with gcc 2.96RH.
> >
> > [1] It's the standard compiler for both RHL and Mandrake
> 
> I find this rather unhelpful, as it really doesn't answer the question.
> I understand that as a RH representative, you stand by the decision to use
> gcc 2.96, but no matter how wonderful you think it is, the MySQL
> directions explicitly state not to use it, because "Several of our users
> have reported random crashes and table corruption with MySQL binaries
> compiled with gcc 2.96 on the x86 Linux platform." 

None of these have been brought to our attention or documented. Note
that the official ones use a different, statically linked version of
glibc as well as a very old compiler.

>  Has this issue been resolved?  If so, the directions should be
> updated.  If not, I question the helpfulness of advising people to
> ignore the directions.
> 
> I have no interest in the arguments over whether 2.96 was a good idea or
> not, and I take you at your word when you tout the merits of 2.96, but
> that's not the question here.  The question is which compiler to use for
> *mysql*.  The list of people with stable copies of mysql compiled with gcc
> 2.95.3 is long, and it includes people who had problems with copies
> compiled with 2.96.  Personally, I have no idea whether that's because of
> a flaw in gcc 2.96 or in mysql, but it's a moot point until it's found and
> fixed.  Has it been?

No issues in interaction between gcc 2.96RH and MySQL has been
identified since just after the release of Red Hat Linux 7 a year and
a half ago (we submitted fixes for these to the MySQL team, and they
have been included since the 3.23.midtwenties or so).

A generic issue with MySQL is that MySQL seems more stable when
compiled with "-fno-rtti -fno-exceptions" regardless of the compiler
version. The newest rpms at http://people.redhat.com/teg/mysql/ are
compiled that way, older ones aren't. Still haven't heard of any
issues or experience any myself.

FWIW, I've ran quite a few tests (the regression tests, the benchmarks
etc) with no stability problems whatsoever.
-- 
Trond Eivind Glomsrød
Red Hat, Inc.

---------------------------------------------------------------------
Before posting, please check:
   http://www.mysql.com/manual.php   (the manual)
   http://lists.mysql.com/           (the list archive)

To request this thread, e-mail <[EMAIL PROTECTED]>
To unsubscribe, e-mail <[EMAIL PROTECTED]>
Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php

Reply via email to