This is the mail archive of the
mailing list for the GCC project.
Re: criteria.html open issues
- To: dan at cgsoftware dot com
- Subject: Re: criteria.html open issues
- From: Marc Espie <espie at quatramaran dot ens dot fr>
- Date: Wed, 30 May 2001 20:58:30 +0200
- Cc: gcc at gcc dot gnu dot org
- Organization: Ecole Normale Superieure (quatramaran)
- References: <200105290308.f4T38ef161237@saturn.cs.uml.edu>
In article <firstname.lastname@example.org> you write:
>This involves knowing the source code.
>So for people who have been maintaining a given package for a while,
>scanning for obvious aliasing problems isn't that big a deal, since
>they're in all the code anyway fixing random reported bugs.
You do seem to imply that aliasing problems are obvious. They are not.
The obvious ones have been taken care of ages ago. What was left was
definitely non obvious.
Very low-level machine code. Language interpreters.
This does not strike me as places where bugs like this would be very obvious.
And you seem to think that people know the code they maintain thoroughly.
Well, I've seen some fairly interesting discussions on this mailing-list in
the past, about some transient bugs, and some very interesting piece of gcc
called `reload', and some apparent difficulty in nailing down exactly what
it was supposed to do under some circumstances.
That, from the very people that maintain gcc.
Do you really think this kind of situation is that uncommon ?
Worse, aliasing bugs only occur in optimized code, and in somewhat
global situations. I'm hardly put to classify bugs that would be harder