The open() system call in f8 really broken...

[Simo Sorce]

On Thu, 2007-08-16 at 17:58 +0200, Oliver Falk wrote:
> On 08/16/2007 05:54 PM, Simo Sorce wrote:
> > On Thu, 2007-08-16 at 11:20 -0400, Steve Dickson wrote:
> >> Oliver Falk wrote:
> >>>> *Anything* is better than having glibc calling abort()... imho...
> >>> Wouldn't most programs - I can think of some big commercial DB - then
> >>> abort as well?
> >> Letting commercial DB abort is much different than having glibc abort
> >> They are making the decision on what to do, not glibc. Leaving those
> >> types of decisions in the handles of the apps is much better than
> >> having glibc playing God.. imho...
> > 
> > +1 not that I am a big fan of proprietary software (not at all), but
> > aborting a DB (or any other software that manages complex structures)
> > with the risk of corrupting the DB is definitely WRONG.
> 
> I'm also not a big fan... But if you run proprietary software it's
> usually supported under specific distributions. And the vendor will hit
> that bug and will not support that distro until they have fixed the code.

True, if that is a compile time error, but this is runtime, and is
something that may happen just maybe in a very little exercised code
path like an exotic error handling routine.

> Also, as I already mentioned. You don't get EL6 (or whatever), copy your
> software to the new machine and take it into production... You'll
> usually test your software first, don't you?

Fair 'nuf.

Simo.