This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: gcc 3.2.1 schedule?
- From: prj at po dot cwru dot edu (Paul Jarc)
- To: gcc at gcc dot gnu dot org
- Date: Thu, 31 Oct 2002 16:39:30 -0500
- Subject: Re: gcc 3.2.1 schedule?
- Organization: What did you have in mind? A short, blunt, human pyramid?
- References: <200210312024.VAA07815@faui11.informatik.uni-erlangen.de>
Ulrich Weigand <weigand@immd1.informatik.uni-erlangen.de> wrote:
> Paul Jarc wrote:
>> Is there some reason why 8146 hasn't been made high-priority? It's a
>> regression, right?
>
> [The problem is bootstrapping gcc 2.95.3 with gcc 3.2 gives
> comparison failures.]
>
> Are you sure this signifies a bug in gcc 3.2, as opposed to
> a bug in gcc 2.95.3 that simply gets exposed by being compiled
> with 3.2?
No. I was just going by the statement (which I can't find ATM) that a
bootstrap comparison failure usually indicates a bug in the compiler
you start with.
> Does the bootstrap work when stage 1 is built with
> -fno-strict-aliasing?
Yes: I used an existing 3.2 build and added -fno-strict-aliasing to
STAGE1_CFLAGS for a 2.95.3 build, and there were no errors this time.
Does this mean it's definitely not a 3.2 bug?
>> (It's also the reason I still use 2.95.3; I'm not
>> happy using a compiler that I know to produce wrong code.)
>
> That assumes gcc 2.95.3 is not known to produce wrong code ;-)
No, it assumes *I* don't know 2.95.3 to produce wrong code. :)
Anyway, GNATS seems to have many more wrong-code reports against 3.*
than against 2.95.3.
paul