GCC silently miscompiling code on x86

GCC silently miscompiling code on x86

[Greg Schafer]

Hello gcc hackers!

This is very bad for i686. Many continue to get hit by this. H J Lu has
proposed some patches but they remain unreviewed. Could somebody kindly
take a look at these and get this sorted?

http://gcc.gnu.org/ml/gcc-patches/2002-08/msg01170.html

http://gcc.gnu.org/ml/gcc-patches/2002-07/msg01561.html

Thanks in advance.

Greg

Re: GCC silently miscompiling code on x86

[H. J. Lu]

On Fri, Aug 23, 2002 at 08:54:00AM +1000, Greg Schafer wrote:
> Hello gcc hackers!
> 
> This is very bad for i686. Many continue to get hit by this. H J Lu has
> proposed some patches but they remain unreviewed. Could somebody kindly
> take a look at these and get this sorted?
> 
> http://gcc.gnu.org/ml/gcc-patches/2002-08/msg01170.html
> 
> http://gcc.gnu.org/ml/gcc-patches/2002-07/msg01561.html
> 

I am enclosing an update for this one. There is no regression on
Linux/i386/i586/i686.


H.J.

Re: GCC silently miscompiling code on x86

[Richard Henderson]

On Fri, Aug 23, 2002 at 08:54:00AM +1000, Greg Schafer wrote:
> This is very bad for i686. Many continue to get hit by this. H J Lu has
> proposed some patches but they remain unreviewed. Could somebody kindly
> take a look at these and get this sorted?

As far as I can tell, HJ's patch doesn't work.  Certainly
it didn't change the output for the given test case at all.

I have a different patch that I started some time ago before I
left on vacation that fixes this bug, as well as the variable
sized structure copy bug as seen in execute/20020412-1.c.


r~

Re: GCC silently miscompiling code on x86

[Greg Schafer]

On Thu, Aug 29, 2002 at 12:04:43PM -0700, Richard Henderson wrote:
> On Fri, Aug 23, 2002 at 08:54:00AM +1000, Greg Schafer wrote:
> > This is very bad for i686. Many continue to get hit by this. H J Lu has
> > proposed some patches but they remain unreviewed. Could somebody kindly
> > take a look at these and get this sorted?
> 
> As far as I can tell, HJ's patch doesn't work.  Certainly
> it didn't change the output for the given test case at all.

That was not my experience at all. HJ's patch here:-

http://gcc.gnu.org/ml/gcc-patches/2002-08/msg01401.html

fixed 20020412-1.c for me. And this one:-

http://gcc.gnu.org/ml/gcc-patches/2002-08/msg01170.html

fixes the test case in GNATS PR 7591.

I am compiling everything at -march=i686

> I have a different patch that I started some time ago before I
> left on vacation that fixes this bug, as well as the variable
> sized structure copy bug as seen in execute/20020412-1.c.

Ok, cool, I would love to test it out. Is it ready for consumption?

Greg

Re: GCC silently miscompiling code on x86

[Richard Henderson]

On Fri, Aug 30, 2002 at 08:03:07AM +1000, Greg Schafer wrote:
> > I have a different patch that I started some time ago before I
> > left on vacation that fixes this bug, as well as the variable
> > sized structure copy bug as seen in execute/20020412-1.c.
> 
> Ok, cool, I would love to test it out. Is it ready for consumption?

Yes, committed to mainline today.

   http://gcc.gnu.org/ml/gcc-patches/2002-08/msg01715.html



r~

Re: GCC silently miscompiling code on x86

[H. J. Lu]

On Thu, Aug 29, 2002 at 12:04:43PM -0700, Richard Henderson wrote:
> On Fri, Aug 23, 2002 at 08:54:00AM +1000, Greg Schafer wrote:
> > This is very bad for i686. Many continue to get hit by this. H J Lu has
> > proposed some patches but they remain unreviewed. Could somebody kindly
> > take a look at these and get this sorted?
> 
> As far as I can tell, HJ's patch doesn't work.  Certainly
> it didn't change the output for the given test case at all.

I guess people may tell you otherwise.

> 
> I have a different patch that I started some time ago before I
> left on vacation that fixes this bug, as well as the variable
> sized structure copy bug as seen in execute/20020412-1.c.
> 

I am looking forward to it.


H.J.