GCC 4.0.1 Status

GCC 4.0.1 Status

[Mark Mitchell]

As of now, please consider the GCC 4.0 branch completely frozen for all 
changes, including documentation, testsuite, etc.

However, I'm not entirely happy with the state of the compiler from the 
point of view of doing a release.

There are several PRs for wrong-code in 4.0.1 that seem potentially as 
serious as the KDE thing.  In order from most serious to least:

* PR 22043, which involves a failure to initialize fields in automatic 
structures.  The patch has been applied to the 4.0 branch, but the 
target milestone still says 4.0.1.  Bugmasters, is that just a mistake?

* PR 22000, in which we throw away volatile reads.  That's a serious 
problem for embedded targets.

* PR 22051, which involves wrong-code generation for pointer comparisons 
on one of our primary platforms.

* PR 21985, in which we are mis-folding expressions involving pointer 
arithmetic.

I'll wander over to the summit, and see whose brains are available for 
picking about these things.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304

Re: GCC 4.0.1 Status

[Andrew Pinski]

On Jun 23, 2005, at 10:08 AM, Mark Mitchell wrote:

> * PR 22043, which involves a failure to initialize fields in automatic 
> structures.  The patch has been applied to the 4.0 branch, but the 
> target milestone still says 4.0.1.  Bugmasters, is that just a 
> mistake?

Yes this was a mistake.

> * PR 22000, in which we throw away volatile reads.  That's a serious 
> problem for embedded targets.

And I think it is still latent on the mainline too, just harder to 
expose because of
the changes to what passes are run and when.  I found a testcase for 
the mainline too.

> * PR 22051, which involves wrong-code generation for pointer 
> comparisons on one of our primary platforms.

There is a missing cast which causes this but I have not looked into who
is causing this, it might be fold.


> * PR 21985, in which we are mis-folding expressions involving pointer 
> arithmetic.

And this is fixed on the mainline already.

-- Pinski

Re: GCC 4.0.1 Status

[Mark Mitchell]

Andrew Pinski wrote:

> 
>> * PR 21985, in which we are mis-folding expressions involving pointer 
>> arithmetic.
> 
> And this is fixed on the mainline already.

The same patch does fix the problem on the branch; I'm now running a 
bootstrap/test cycle using that patch.

Thanks,

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304