segv during Wine-980104 build

segv during Wine-980104 build

Post by Howard Goldste » Sat, 17 Jan 1998 04:00:00



Trying to build Wine-980104 from FreeBSD current ports

cc -c -O -Wall -D__WINE__  -I../include -I../include
-I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
./shellord.c: In function `SHELL32_183':
./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from incompatible pointer type
./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
cc: Internal compiler error: program cc1 got fatal signal 11

...

cally:/usr/ports/emulators/wine# gcc -v
gcc version 2.7.2.1
cally:/usr/ports/emulators/wine# uname -a

cally:/usr/ports/emulators/wine# swapinfo -k
Device      1K-blocks     Used    Avail Capacity  Type
/dev/sd0s3b     73728    13120    60544    18%    Interleaved
/dev/vn0b       65536    13044    52428    20%    Interleaved
Total          139136    26164   112972    19%

Any thoughts?

--
Judging by who's against it, NoCeM must be a good thing     http://www.cm.org

 
 
 

segv during Wine-980104 build

Post by Douglas Ridgw » Sun, 18 Jan 1998 04:00:00


: Trying to build Wine-980104 from FreeBSD current ports

: cc -c -O -Wall -D__WINE__  -I../include -I../include
: -I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
: ./shellord.c: In function `SHELL32_183':
: ./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from incompatible pointer type
: ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
: cc: Internal compiler error: program cc1 got fatal signal 11

Oh good, I thought it was just me. I had the same problem on Red Hat Linux
5.0.

Well, by definition, it must be a gcc bug. I found it quite repeatable, but
in trying to track it down a bit, I found that even very slight
changes caused it to disappear.  (Eg, changing level of optimization,
or removing -Wall or -fPIC, or even renaming the source file!)

The workaround is to just compile that one file by hand. Removing
optimization was sufficient for me, and then I could just restart the
make.

doug.



 
 
 

segv during Wine-980104 build

Post by James Moo » Sun, 18 Jan 1998 04:00:00


: Trying to build Wine-980104 from FreeBSD current ports

: cc -c -O -Wall -D__WINE__  -I../include -I../include
: -I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
: ./shellord.c: In function `SHELL32_183':
: ./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from incompatible pointer type
: ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
: cc: Internal compiler error: program cc1 got fatal signal 11

: ...

: cally:/usr/ports/emulators/wine# gcc -v
: gcc version 2.7.2.1
: cally:/usr/ports/emulators/wine# uname -a

: cally:/usr/ports/emulators/wine# swapinfo -k
: Device      1K-blocks     Used    Avail Capacity  Type
: /dev/sd0s3b     73728    13120    60544    18%    Interleaved
: /dev/vn0b       65536    13044    52428    20%    Interleaved
: Total          139136    26164   112972    19%

: Any thoughts?

You may have bad memory, or perhaps a Cyrix CPU with a badly matched
motherboard. In any case, it's likely a hardware problem. If you type make
again, it may not fail in the same place.

For the signal 11 FAQ, try http://www.bitwizard.nl/sig11/

james

 
 
 

segv during Wine-980104 build

Post by plipp.. » Sun, 18 Jan 1998 04:00:00




>: Trying to build Wine-980104 from FreeBSD current ports

----- clip -----

Quote:>: Any thoughts?
>You may have bad memory, or perhaps a Cyrix CPU with a badly matched
>motherboard. In any case, it's likely a hardware problem. If you type make
>again, it may not fail in the same place.
>For the signal 11 FAQ, try http://www.bitwizard.nl/sig11/

It's also a good idea to check RedHat5.0 errata. There was one part about
sig11 error, that was fixed installing special glibc-gcc. (At least with
Cyrix CPU:s)

url for errata:

http://www.redhat.com/support/docs/rhl/rh50-errata-general.html

PEte
--
__________________________________________________________________________
| "Numbers suck!    There's like  | Petri Lipponen, physics student, HUT |
| too many of them, or something" | tel. +358-50-546 9699                |

 
 
 

segv during Wine-980104 build

Post by Marcus Meissn » Sun, 18 Jan 1998 04:00:00




>Trying to build Wine-980104 from FreeBSD current ports

>cc -c -O -Wall -D__WINE__  -I../include -I../include
>-I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
>./shellord.c: In function `SHELL32_183':
>./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from
>incompatible pointer type
>./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from
>incompatible pointer type
>cc: Internal compiler error: program cc1 got fatal signal 11

I had the same problem, but it went away. Since it always crashes
at the same line and only while compiling WINE (not while compiling
the more complex linux kernel) I suspect a gcc bug.

Marcus
--
   <URL:http://wwwcip.informatik.uni-erlangen.de/~msmeissn/>

 
 
 

segv during Wine-980104 build

Post by Howard Goldste » Sun, 18 Jan 1998 04:00:00




 : : Trying to build Wine-980104 from FreeBSD current ports
...
 : : ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
 : : cc: Internal compiler error: program cc1 got fatal signal 11
 :
 : Oh good, I thought it was just me. I had the same problem on Red Hat Linux
 : 5.0.

(Here on freebsd 2.2.5r and an intel 586/133...)

 :
 : Well, by definition, it must be a gcc bug. I found it quite repeatable, but
 : in trying to track it down a bit, I found that even very slight
 : changes caused it to disappear.  (Eg, changing level of optimization,
 : or removing -Wall or -fPIC, or even renaming the source file!)
 :
 : The workaround is to just compile that one file by hand. Removing
 : optimization was sufficient for me, and then I could just restart the
 : make.

That worked for me too.  I tried first zapping -Wall, then upping to
-O2 but those combos also dumped core.  Definitely appears to be a gcc
optimizer bug.  Another kind person who responded suggested a resource
problem.  I cannot rule it out but I do have something like 100mb of
VM free which in the immortal words of Gill Bates "ought to be enough
for anyone."

Thank you Doug, and thanks also to the kind folks who responded
elsewhere in the thread and via email.

--
Judging by who's against it, NoCeM must be a good thing     http://www.cm.org

 
 
 

segv during Wine-980104 build

Post by Kaleb S. KEITHLE » Sun, 18 Jan 1998 04:00:00





> >Trying to build Wine-980104 from FreeBSD current ports

> >cc -c -O -Wall -D__WINE__  -I../include -I../include
> >-I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
> >./shellord.c: In function `SHELL32_183':
> >./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from
> >incompatible pointer type
> >./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from
> >incompatible pointer type
> >cc: Internal compiler error: program cc1 got fatal signal 11

> I had the same problem, but it went away. Since it always crashes
> at the same line and only while compiling WINE (not while compiling
> the more complex linux kernel) I suspect a gcc bug.

Interesting. On my 2.2.2R system I had no problem compiling, although I
had to fix the "build" utility to emit .ascii instead of .string. But
after all that ld will fall down trying to do the final link of wine.

I've seen ld fall down before trying to build a fully debuggable X
server -- it grows to consume all available swap (I have 100+ Meg of
swap) and finally exits after failing to allocate more memory. It's
doing the same thing with wine, i.e. running out of memory.

I've pulled over the source to the 2.2.5 ld and the 3.0-SNAP ld, and
they're all the same.

I dunno, when there's too many symbols ld goes into an infinite loop
somewhere trying to allocate memory. I've given it a cursory look, but
nothing jumps out. Perhaps someone who knows link-editors better than I
can figure it out. :-(

--

Kaleb S. KEITHLEY

 
 
 

segv during Wine-980104 build

Post by Scott Lo » Mon, 19 Jan 1998 04:00:00






> : : Trying to build Wine-980104 from FreeBSD current ports
> ...
> : : ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
> : : cc: Internal compiler error: program cc1 got fatal signal 11
> :

On FreeBSD, try making it with gmake explicitly, instead of the stock make.
That solved my problem.  If you're making it from the ports directory, add
a line like "USE_GMAKE= yes" in the Makefile.

btw, this segfault doesn't happen with egcs 1.0.1. so.....

Scott

 
 
 

segv during Wine-980104 build

Post by Ollivier Robe » Mon, 19 Jan 1998 04:00:00


    [courtesy cc of this posting sent to cited author via email]



> Speaking of egcs do you know if there've been any attempts to build
> the kernel with it?  Also wonder if any performance wins can be had by
> compiling heavy hitter apps under it, like X or INN...

You can build kernels just fine with it and with pgcc as well (pgcc is now
based on egcs). I've built and run several kernel compiled with -O2/-O3
and even -O6 (& -mcpu=amdk6).

I've not run any benchmark yet.

egcs is faster at compiling and generate much better code than 2.7.2. I've
made an ELF version as well to experiment with the ELF support.
--

Usenet C*Historique

 
 
 

segv during Wine-980104 build

Post by Dr Bill C Rieme » Tue, 20 Jan 1998 04:00:00








>> : : Trying to build Wine-980104 from FreeBSD current ports
>> ...
>> : : ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
>> : : cc: Internal compiler error: program cc1 got fatal signal 11

This is a problem with 'gcc' I've seen under SunOS, Solaris, and Linux for
both sparc and intel platforms, compiling perl5.004, Wine, and several other
applications.  Usually the work around is just to recompile the effected
routine with a different optimization level.  Or in one case I had to use
Sun's compiler.  But I haven't been ambitious enough to figure out who to
report gcc 2.7.2.3 bugs to, so I've just been using the work around.

                                Bill

--

  104D White Street  ' \ . / `  ///'V( (      http://feynman.com
  Eatontown, NJ        :`O':   /'! ::::::-.  Free Speech Online!
  07724 USA            o/U\o  '  !.`::::'-'  Phone:(732)544-4928

 
 
 

segv during Wine-980104 build

Post by Marcus Meissn » Tue, 20 Jan 1998 04:00:00




>    [courtesy cc of this posting sent to cited author via email]



>> Speaking of egcs do you know if there've been any attempts to build
>> the kernel with it?  Also wonder if any performance wins can be had by
>> compiling heavy hitter apps under it, like X or INN...

>You can build kernels just fine with it and with pgcc as well (pgcc is now
>based on egcs). I've built and run several kernel compiled with -O2/-O3
>and even -O6 (& -mcpu=amdk6).

Compiling with -O<nr> with nr > 3 has the same effect as compiling with
-O3. (Glancing at grep -w optimize gcc/*.c. The highest check is for -O3.)

Marcus
--
   <URL:http://wwwcip.informatik.uni-erlangen.de/~msmeissn/>

 
 
 

segv during Wine-980104 build

Post by Ollivier Robe » Wed, 21 Jan 1998 04:00:00


    [courtesy cc of this posting sent to cited author via email]



> Compiling with -O<nr> with nr > 3 has the same effect as compiling with
> -O3. (Glancing at grep -w optimize gcc/*.c. The highest check is for -O3.)

In pgcc (derived from egcs), in gcc/config/i386/i386.c you find this:

-=-=-=-=-=-
static void
optimization_options_intel1 (level)
     int level;
{
  if (level > 0)
    {
      flag_opt_reg_use = 2;
      flag_reduce_index_givs = 2;
    }
  if (level >= 2)
    {
      flag_schedule_insns = 2;
      flag_schedule_insns_after_reload = 2;
    }
  if (level >= 3)
    {
      flag_inline_functions = 2;
      flag_jump_back = 2;
      flag_copy_prop = 2;
      flag_compare_elim = 2;
      flag_sftwr_pipe = 2;
      flag_reg_reg_copy_opt = 2;
      /*flag_opt_reg_stack = 2;*//*D*/
      /*flag_loop_after_global = 2;*//*D*/
      flag_peep_spills = 2;
      flag_replace_stack_mem = 2;
      flag_opt_jumps_out = 2;
      flag_replace_mem = 2;
      flag_correct_cse_mistakes = 2;
      flag_push_load_into_loop = 2;
      flag_replace_reload_regs = 2;
      flag_sign_extension_elim = 2;
      flag_lift_stores = 2;
    }

  if (level >= 4)
    {
      /*flag_combine_222 = 2;*/ /*D*/
#ifdef INSN_SCHEDULING
      flag_schedule_insns_after_reload = 2;
      flag_swap_for_agi = 2;
      flag_risc = 2;
      flag_risc_const = 2;
      /*flag_recombine = 2;*//*D*/ /* ??? actually slows down */
      flag_interleave_stack_non_stack = 2;
      flag_schedule_stack_reg_insns = 2;
#endif
    }
  if (level >= 5)
    {
      flag_runtime_lift_stores = 2; /* big space penalty */
      flag_omit_frame_pointer = 2;
#ifdef INSN_SCHEDULING
      /*flag_schedule_insns = 2;*/ /* hurts performance! */
#endif
    }
  if (level >= 6)
    {
      flag_all_mem_givs = 2;
      flag_do_offload = 2;
      flag_risc_mem_dest = 2;
    }
/*  if (level >= 7)
    {
      flag_unroll_loops = 2;
      flag_sibling_call = 2;
    }*/

Quote:}

-=-=-=-=-=-
--

Usenet C*Historique
 
 
 

segv during Wine-980104 build

Post by Donn Mille » Sat, 24 Jan 1998 04:00:00



Quote:> Trying to build Wine-980104 from FreeBSD current ports

> cc -c -O -Wall -D__WINE__  -I../include -I../include
> -I. -I. -I/usr/X11R6/include -o shellord.o ./shellord.c
> ./shellord.c: In function `SHELL32_183':
> ./shellord.c:522: warning: passing arg 5 of `FormatMessage32A' from incompatible pointer type
> ./shellord.c:522: warning: passing arg 7 of `FormatMessage32A' from incompatible pointer type
> cc: Internal compiler error: program cc1 got fatal signal 11

> ...

Mine crashes in the exact same location, so it can't be a memory problem,
except I'm running 3.0-980117-SNAP.  Maybe a bug in cc1?

> cally:/usr/ports/emulators/wine# gcc -v
> gcc version 2.7.2.1
> cally:/usr/ports/emulators/wine# uname -a

> cally:/usr/ports/emulators/wine# swapinfo -k
> Device      1K-blocks     Used    Avail Capacity  Type
> /dev/sd0s3b     73728    13120    60544    18%    Interleaved
> /dev/vn0b       65536    13044    52428    20%    Interleaved
> Total          139136    26164   112972    19%

> Any thoughts?

> --
> Judging by who's against it, NoCeM must be a good thing     http://www.cm.org

 
 
 

1. Wine,wine,wine - I need help with wine

Help
I down loaded and installed Wine to Linux 2.0.22 and Xwindows (XFree).
I think it's installed correctly.  I am getting a 'drive not found/path
not found' error.  I have my wine.conf in the right place (I think).
I'm using Win95 and hda3 and Linux on hda1.  I need a reality check.  Is
it hard to set up Wine?  Does someone have a working wine.conf file that
might guide me to a working wine.   Wine,wine,wine and beer. Ok I'll
stop wineing

Thanks
George

--

--
**************************************************
*                                                *
*   -----(((( Every Dog Has His Day ))))-----    *
*       ( to stay on the porch and bark )        *
*                                                *
*                George Dunham                   *
*            "Crappy Computers Ink"              *

*     http://www.sni.net/~gdunham/gdunh.htm      *
*    arf             |\__/\                      *
*        arf arf     / oo \_                     *
*             arf    | o  | \                    *
*                arf  \()/   \   )               *
**************************************************

2. Ping across router problem

3. Got SEGV during link

4. Unable to ftp when logon in tcsh

5. page access during SEGV on linux ?

6. mail reciving reset, and possible file corruption, please HELP

7. wine: port does not build

8. fax software for linux??

9. Latest Wine build on FreeBSD 2.1.0-R ???

10. Building Wine

11. KDevelop and Wine, MFC, C++, Wine stability

12. WINE - wine.conf

13. Wine=NT like DOSemu=DOS (was Re: wctype.h (was: Wine release 971221))