[LinuxBIOS] romcc - Internal compiler error: Cannot find block dominated ....

[Doug Bell]

0x8360d58 phi        Internal compiler error: Cannot
find block dominated by 0x82ba9a8

Any ideas on workarounds / fixes here? Is am I running
out of registers or some other problem?

__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com

[ron minnich]

On 9/26/05, Doug Bell  wrote:
> 0x8360d58 phi Internal compiler error: Cannot
> find block dominated by 0x82ba9a8
>
> Any ideas on workarounds / fixes here? Is am I running
> out of registers or some other problem?

can you tell us more about the environment,
uname -a
distro, and linuxbios target?

ron

[Doug Bell]

It seems to not be related to register allocation, how
many layers of functions,  or code size, but to nested
if statements or variations thereof, e.g. if (
(condition a) && ( condition b) ) ....

/home/dbell # uname -a
Linux Doug 2.6.9 #1 Tue Jan 11 17:26:41 UTC 2005 i686
Intel(R) Pentium(R) 4 CPU 1500MHz GenuineIntel
GNU/Linux

Target is via epia-sp, under development but close to
an alpha release with some open items.

BTW, is there any documentation regarding what
limitations romcc has (e.g. multidimensional arrays
not allowed, etc )?

Thanks, Doug

--- ron minnich  wrote:
> On 9/26/05, Doug Bell  wrote:
> > 0x8360d58 phi Internal compiler error: Cannot
> > find block dominated by 0x82ba9a8
> >
> > Any ideas on workarounds / fixes here? Is am I
>
> running
> > out of registers or some other problem?
>
> can you tell us more about the environment,
> uname -a
> distro, and linuxbios target?
>
> ron
> >
>
> LinuxBIOS mailing list
> [e-mail]
> http://www.openbios.org/mailman/listinfo/linuxbios

__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com

[Doug Bell]

I had one version of code where it failed on a line
like:
if (speed==100) speed=0;

but problem disappeared if I did :
if (speed==100) speed2=0;

This may have been a factor in the first reporting of
the problem as well.

-dB

--- Doug Bell  wrote:
> It seems to not be related to register allocation,
> how
> many layers of functions,  or code size, but to
> nested
> if statements or variations thereof, e.g. if (
> (condition a) && ( condition b) ) ....
>
> /home/dbell # uname -a
> Linux Doug 2.6.9 #1 Tue Jan 11 17:26:41 UTC 2005
> i686
> Intel(R) Pentium(R) 4 CPU 1500MHz GenuineIntel
> GNU/Linux
>
> Target is via epia-sp, under development but close
> to
> an alpha release with some open items.
>
> BTW, is there any documentation regarding what
> limitations romcc has (e.g. multidimensional arrays
> not allowed, etc )?
>
> Thanks, Doug
>
> --- ron minnich  wrote:
> > On 9/26/05, Doug Bell  wrote:
> > > 0x8360d58 phi Internal compiler error: Cannot
> > > find block dominated by 0x82ba9a8
> > >
> > > Any ideas on workarounds / fixes here? Is am I
> >
> > running
> > > out of registers or some other problem?
> >
> > can you tell us more about the environment,
> > uname -a
> > distro, and linuxbios target?
> >
> > ron
> > >
> >
> > LinuxBIOS mailing list
> > [e-mail]
> > http://www.openbios.org/mailman/listinfo/linuxbios
>
> __________________________________ 
> Yahoo! Mail - PC Magazine Editors' Choice 2005 
> http://mail.yahoo.com

__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com

[Doug Bell]

Beyond that, there was one version in which I simply
switched the order of declaration of variables to make
it compile:

// unsigned char row,col; // broken
unsigned char col,row; // works

In another test run........
//if ( total < c) total+=div; // failed

if ( total < c) subtotal+=div; // was ok
total=subtotal;

Ron / Eric  - What is the best way to proceed on this?
What info would you want? I don't think I'm doing
anything bizarre, maybe 4 chars declared, no deep
levels of functions.

--- Doug Bell  wrote:
> I had one version of code where it failed on a line
> like:
> if (speed==100) speed=0;
>
> but problem disappeared if I did :
> if (speed==100) speed2=0;
>
> This may have been a factor in the first reporting
> of
> the problem as well.
>
> -dB
>
> --- Doug Bell  wrote:
> > It seems to not be related to register allocation,
> > how
> > many layers of functions,  or code size, but to
> > nested
> > if statements or variations thereof, e.g. if (
> > (condition a) && ( condition b) ) ....
> >
> > /home/dbell # uname -a
> > Linux Doug 2.6.9 #1 Tue Jan 11 17:26:41 UTC 2005
> > i686
> > Intel(R) Pentium(R) 4 CPU 1500MHz GenuineIntel
> > GNU/Linux
> >
> > Target is via epia-sp, under development but close
> > to
> > an alpha release with some open items.
> >
> > BTW, is there any documentation regarding what
> > limitations romcc has (e.g. multidimensional
>
> arrays
> > not allowed, etc )?
> >
> > Thanks, Doug
> >
> > --- ron minnich  wrote:
> > > On 9/26/05, Doug Bell  wrote:
> > > > 0x8360d58 phi Internal compiler error: Cannot
> > > > find block dominated by 0x82ba9a8
> > > >
> > > > Any ideas on workarounds / fixes here? Is am I
> > >
> > > running
> > > > out of registers or some other problem?
> > >
> > > can you tell us more about the environment,
> > > uname -a
> > > distro, and linuxbios target?
> > >
> > > ron
> > > >
> > >
> > > LinuxBIOS mailing list
> > > [e-mail]
>
> http://www.openbios.org/mailman/listinfo/linuxbios
> > __________________________________ 
> > Yahoo! Mail - PC Magazine Editors' Choice 2005 
> > http://mail.yahoo.com
>
> __________________________________ 
> Yahoo! Mail - PC Magazine Editors' Choice 2005 
> http://mail.yahoo.com

__________________________________ 
Yahoo! Mail - PC Magazine Editors' Choice 2005 
http://mail.yahoo.com

Doug Bell  writes:
> 0x8360d58 phi        Internal compiler error: Cannot
> find block dominated by 0x82ba9a8
>
> Any ideas on workarounds / fixes here?

Try commenting out code to isolate the problem or remove.
> Is am I running out of registers or some other problem?

Unless something has found a way to miscompile romcc it
is a bug in the internal romcc logic.  Bugs of that sort are
pretty rare and almost always they break the compile instead
of giving you bad code.

This feels like something was optimized away and not everything
was properly updated.  Either that or there is a spaghetti
mess of goto's somewhere in that chunk of code.

Eric