gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Pretty scary, this one (using gcc-2.95pre):

[hirlam@moene ref]$ gdb hlprog32.x core 
GNU gdb 4.17.0.4 with Linux/x86 hardware watchpoint and FPU support
Copyright 1998 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you
are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for
details.
This GDB was configured as "i386-redhat-linux"...
Core was generated by `/usr/scratch/hirlam/ref/hlprog32.x'.
Program terminated with signal 11, Segmentation fault.
#0  init (argc=1, argv=0xbffff88c, envp=0xbffff894)
    at ../sysdeps/unix/sysv/linux/init-first.c:70
../sysdeps/unix/sysv/linux/init-first.c:70: No such file or directory.

For a program that plain works under the compilers from the trunk -
errmmm, oh yes, I changed this one in f/com.c for those:

#define FFECOM_sizeMAXSTACKITEM 0
#ifndef FFECOM_sizeMAXSTACKITEM
#define FFECOM_sizeMAXSTACKITEM 32*1024 /* Keep user-declared things
                                           larger than this # bytes
                                           off stack if possible. */
#endif

[ the first line is mine ]

What this does is make the compiler generate code to allocate all local
arrays on the stack, regardless of size.

Which means that the error I get is still pretty scary:  A 45 Mbyte
working set program that cannot be linked successfully on RH Linux 5.2 -
brrr :-(

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Sylvain Pion]

On Sun, Jun 13, 1999 at 07:35:30PM +0200, Toon Moene wrote:
> What this does is make the compiler generate code to allocate all local
> arrays on the stack, regardless of size.

If you go beyond the default stack size (8MB), you'll get a sig11.
Did you try running it with "unlimit stacksize" ?
For example, just run the following:

int main()
{
  int a[1<<21];
  return a[0];
}

-- 
Sylvain

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Sylvain Pion wrote:

> On Sun, Jun 13, 1999 at 07:35:30PM +0200, Toon Moene wrote:
> > What this does is make the compiler generate code to allocate all local
> > arrays on the stack, regardless of size.
> 
> If you go beyond the default stack size (8MB), you'll get a sig11.
> Did you try running it with "unlimit stacksize" ?
> For example, just run the following:

Ah, sorry, I see that my mail was less than clear.  Try the following
Fortran source:

      common /com/ a(20000000)
      a(1) = 1.
      print*, a(1)
      call sub
      end
      subroutine sub
      dimension b(10000000)
      b(1) = 2.
      print*, b(1)
      end

This should pass (as it does), even if b were allocated on the stack
with the following stack size limit:

[toon@moene g77-bugs]$ ulimit -s
65536

However, as distributed, g77 will allocate b statically, because it is
larger than 32 Kbyte;  I can simulate that with the compiler as I
modified it by compiling it thusly:

[toon@moene g77-bugs]$ /usr/rel/bin/g77 -fno-automatic s.f
[toon@moene g77-bugs]$ ./a.out
Segmentation fault (core dumped)

For completeness:

[toon@moene g77-bugs]$ /usr/rel/bin/g77 -v
g77 version gcc-2.95 19990613 (prerelease) (from FSF-g77 version 0.5.25
19990526 (prerelease))
Driving: /usr/rel/bin/g77 -v -c -xf77-version /dev/null -xnone
Reading specs from /usr/rel/lib/gcc-lib/i686-pc-linux-gnu/gcc-2.95/specs
gcc version gcc-2.95 19990613 (prerelease)
... etc ...

As I said, stock Red Hat Linux 5.2.

Cheers,

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Toon Moene wrote:

False Alarm [ It's the Swap Space, Stupid ! ]

> Ah, sorry, I see that my mail was less than clear.  Try the following
> Fortran source:
> 
>       common /com/ a(20000000)
>       a(1) = 1.
>       print*, a(1)
>       call sub
>       end
>       subroutine sub
>       dimension b(10000000)
>       b(1) = 2.
>       print*, b(1)
>       end

4 * (20 000 000 + 10 000 000) = 120 000 000 which is just a little bit
less than 128 Mbyte, which is RAM + Swap on my machine.  Given that I
also have three Netscape windows open - and you see the connection.

Changing 20 000 000 to 15 000 000 makes it work.

My Apologies,

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Sylvain Pion]

On Sun, Jun 13, 1999 at 07:35:30PM +0200, Toon Moene wrote:
> What this does is make the compiler generate code to allocate all local
> arrays on the stack, regardless of size.

If you go beyond the default stack size (8MB), you'll get a sig11.
Did you try running it with "unlimit stacksize" ?
For example, just run the following:

int main()
{
  int a[1<<21];
  return a[0];
}

-- 
Sylvain

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Sylvain Pion wrote:

> On Sun, Jun 13, 1999 at 07:35:30PM +0200, Toon Moene wrote:
> > What this does is make the compiler generate code to allocate all local
> > arrays on the stack, regardless of size.
> 
> If you go beyond the default stack size (8MB), you'll get a sig11.
> Did you try running it with "unlimit stacksize" ?
> For example, just run the following:

Ah, sorry, I see that my mail was less than clear.  Try the following
Fortran source:

      common /com/ a(20000000)
      a(1) = 1.
      print*, a(1)
      call sub
      end
      subroutine sub
      dimension b(10000000)
      b(1) = 2.
      print*, b(1)
      end

This should pass (as it does), even if b were allocated on the stack
with the following stack size limit:

[toon@moene g77-bugs]$ ulimit -s
65536

However, as distributed, g77 will allocate b statically, because it is
larger than 32 Kbyte;  I can simulate that with the compiler as I
modified it by compiling it thusly:

[toon@moene g77-bugs]$ /usr/rel/bin/g77 -fno-automatic s.f
[toon@moene g77-bugs]$ ./a.out
Segmentation fault (core dumped)

For completeness:

[toon@moene g77-bugs]$ /usr/rel/bin/g77 -v
g77 version gcc-2.95 19990613 (prerelease) (from FSF-g77 version 0.5.25
19990526 (prerelease))
Driving: /usr/rel/bin/g77 -v -c -xf77-version /dev/null -xnone
Reading specs from /usr/rel/lib/gcc-lib/i686-pc-linux-gnu/gcc-2.95/specs
gcc version gcc-2.95 19990613 (prerelease)
... etc ...

As I said, stock Red Hat Linux 5.2.

Cheers,

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html

Re: gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Toon Moene wrote:

False Alarm [ It's the Swap Space, Stupid ! ]

> Ah, sorry, I see that my mail was less than clear.  Try the following
> Fortran source:
> 
>       common /com/ a(20000000)
>       a(1) = 1.
>       print*, a(1)
>       call sub
>       end
>       subroutine sub
>       dimension b(10000000)
>       b(1) = 2.
>       print*, b(1)
>       end

4 * (20 000 000 + 10 000 000) = 120 000 000 which is just a little bit
less than 128 Mbyte, which is RAM + Swap on my machine.  Given that I
also have three Netscape windows open - and you see the connection.

Changing 20 000 000 to 15 000 000 makes it work.

My Apologies,

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html

gcc-2.95: Program dies with Sig 11 before doing anything - RH Linux 5.2

[Toon Moene]

Pretty scary, this one (using gcc-2.95pre):

[hirlam@moene ref]$ gdb hlprog32.x core 
GNU gdb 4.17.0.4 with Linux/x86 hardware watchpoint and FPU support
Copyright 1998 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you
are
welcome to change it and/or distribute copies of it under certain
conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB.  Type "show warranty" for
details.
This GDB was configured as "i386-redhat-linux"...
Core was generated by `/usr/scratch/hirlam/ref/hlprog32.x'.
Program terminated with signal 11, Segmentation fault.
#0  init (argc=1, argv=0xbffff88c, envp=0xbffff894)
    at ../sysdeps/unix/sysv/linux/init-first.c:70
../sysdeps/unix/sysv/linux/init-first.c:70: No such file or directory.

For a program that plain works under the compilers from the trunk -
errmmm, oh yes, I changed this one in f/com.c for those:

#define FFECOM_sizeMAXSTACKITEM 0
#ifndef FFECOM_sizeMAXSTACKITEM
#define FFECOM_sizeMAXSTACKITEM 32*1024 /* Keep user-declared things
                                           larger than this # bytes
                                           off stack if possible. */
#endif

[ the first line is mine ]

What this does is make the compiler generate code to allocate all local
arrays on the stack, regardless of size.

Which means that the error I get is still pretty scary:  A 45 Mbyte
working set program that cannot be linked successfully on RH Linux 5.2 -
brrr :-(

-- 
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
GNU Fortran: http://world.std.com/~burley/g77.html