guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

I just tried building up the input file by hand that snarf-check-and-output-texi
expects, and running it through snarf-check-and-output-texi to see what
it does.  The output:

$ snarf-check-and-output-texi </tmp/guile-doc

;;; WARNING (no code for module (ice-9 match))
ERROR: unrecognized defmodule argument (:export (snarf-check-and-output-texi))


It does the same thing with an empty input file.  So it appears that
the guile that is built fails to handle snarf-check-and-output-texi,
but I'm not knowledgeable enough to figure out where it is going
astray.

Again, suggestions are welcome.



					Regards,

					Andrew Koenig


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Marius Vollmer]

Andrew Koenig <ark@research.att.com> writes:

> I just tried building up the input file by hand that
> snarf-check-and-output-texi expects, and running it through
> snarf-check-and-output-texi to see what it does.  The output:
> 
> $ snarf-check-and-output-texi </tmp/guile-doc
> 
> ;;; WARNING (no code for module (ice-9 match))
> ERROR: unrecognized defmodule argument (:export (snarf-check-and-output-texi))
That should be something like

  $ GUILE="../pre-inst-guile" ../scripts/snarf-check-and-output-texi

i.e., the script should use the about-to-be-installed version of
Guile, not the one already installed in the system.  

>From your build logs, I see that
/tmp/build-gnu5877/guile-1.6.3/pre-inst-guile is used, which seems
correct.  Does pre-inst-guile work on your system?

> It does the same thing with an empty input file.  So it appears that
> the guile that is built fails to handle snarf-check-and-output-texi,

It might also be that the guile that is built is not used to run
snarf-check-and-output-texi.  Can you try to figure out which is guile
is actually run, and why?

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

Marius> From your build logs, I see that
Marius> /tmp/build-gnu5877/guile-1.6.3/pre-inst-guile is used, which
Marius> seems correct.  Does pre-inst-guile work on your system?

>> It does the same thing with an empty input file.  So it appears that
>> the guile that is built fails to handle snarf-check-and-output-texi,

Marius> It might also be that the guile that is built is not used to
Marius> run snarf-check-and-output-texi.  Can you try to figure out
Marius> which is guile is actually run, and why?

OK ... what's happening is that the right guile is being run, but
the run is not successful:

[europa] GUILE="/tmp/guile/guile-1.6.3/pre-inst-guile" ../scripts/snarf-check-and-output-texi --manual
Segmentation Fault - core dumped

I believe that the executable that ultimately winds up generating the
core file is in libguile/.libs/guile; on that assumption, here's a traceback:


[europa] gdb libguile/.libs/guile core
GNU gdb 5.3
Copyright 2002 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 "sparc-sun-solaris2.8"...
Core was generated by `guile'.
Program terminated with signal 11, Segmentation fault.
Reading symbols from /tmp/guile/guile-1.6.3/libguile/./.libs/libguile.so.12...done.
Loaded symbols for /tmp/guile/guile-1.6.3/libguile/./.libs/libguile.so.12
Reading symbols from /tmp/guile/guile-1.6.3/libguile-ltdl/.libs/libguile-ltdl.so.1...done.
Loaded symbols for /tmp/guile/guile-1.6.3/libguile-ltdl/.libs/libguile-ltdl.so.1
Reading symbols from /usr/lib/libdl.so.1...done.
Loaded symbols for /usr/lib/libdl.so.1
Reading symbols from /usr/lib/libcrypt_i.so.1...done.
Loaded symbols for /usr/lib/libcrypt_i.so.1
Reading symbols from /usr/lib/libsocket.so.1...done.
Loaded symbols for /usr/lib/libsocket.so.1
Reading symbols from /usr/lib/libnsl.so.1...done.
Loaded symbols for /usr/lib/libnsl.so.1
Reading symbols from /usr/lib/libm.so.1...done.
Loaded symbols for /usr/lib/libm.so.1
Reading symbols from /usr/lib/libc.so.1...done.
Loaded symbols for /usr/lib/libc.so.1
Reading symbols from /usr/gnu/lib/libgcc_s.so.1...done.
Loaded symbols for /usr/gnu/lib/libgcc_s.so.1
Reading symbols from /usr/lib/libgen.so.1...done.
Loaded symbols for /usr/lib/libgen.so.1
Reading symbols from /usr/lib/libmp.so.2...done.
Loaded symbols for /usr/lib/libmp.so.2
Reading symbols from /usr/platform/SUNW,Ultra-5_10/lib/libc_psr.so.1...done.
Loaded symbols for /usr/platform/SUNW,Ultra-5_10/lib/libc_psr.so.1
#0  0xff2b1810 in scm_ceval (x=0x48e40, env=0x2974) at eval.c:1986
1986          while (!SCM_NULLP (t.arg1 = SCM_CDR (t.arg1)))


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Marius Vollmer]

Andrew Koenig <ark@research.att.com> writes:

> OK ... what's happening is that the right guile is being run, but
> the run is not successful:
> 
> [europa] GUILE="/tmp/guile/guile-1.6.3/pre-inst-guile" ../scripts/snarf-check-and-output-texi --manual
> Segmentation Fault - core dumped

Ok.  I have just compiled guile-1.6.3 on

  SunOS wiggum 5.8 Generic_108528-15 sun4u sparc SUNW,Ultra-5_10

with GCC 2.95.3 and it works fine.

It looks like you might have used GCC 3.x, is that right?  (Your
binary uses libgcc_s.so which will only happen with GCC 3.x, right?)
We had problems with some new optimizations in GCC on IA32 and there
might be different issues on Sparc.

Could you try compiling without "-O2"?  Also, if convenient, could you
compile with GCC 2.95.x?

Of that fixes it, we will then have to find a better way to make Guile
work on Sparc, of course.  I have the hunch that you known something
about C, Andrew... ;-) so maybe you can tell us what to do.


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

Marius> It looks like you might have used GCC 3.x, is that right?  (Your
Marius> binary uses libgcc_s.so which will only happen with GCC 3.x, right?)
Marius> We had problems with some new optimizations in GCC on IA32 and there
Marius> might be different issues on Sparc.

I did use GCC 3.x

Marius> Could you try compiling without "-O2"?  Also, if convenient, could you
Marius> compile with GCC 2.95.x?

Do I need to do anything beyond deleting "-O2" from all the Makefiles?

Marius> Of that fixes it, we will then have to find a better way to make Guile
Marius> work on Sparc, of course.  I have the hunch that you known something
Marius> about C, Andrew... ;-) so maybe you can tell us what to do.

It may not be easy for me to compile with gcc 2.95 on my machine; I'm not
sure I have a working gcc 2 installation.  I'll give it a try, though.


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Marius Vollmer]

Andrew Koenig <ark@research.att.com> writes:

> Marius> Could you try compiling without "-O2"?  Also, if convenient,
> could you Marius> compile with GCC 2.95.x?
> 
> Do I need to do anything beyond deleting "-O2" from all the Makefiles?

Just doing

  $ make clean
  $ make CFLAGS=-g

should be enough.  You can override the nonessential compiler flags by
setting CFLAGS on the make command line.

> Marius> Of that fixes it, we will then have to find a better way to make Guile
> Marius> work on Sparc, of course.  I have the hunch that you known something
> Marius> about C, Andrew... ;-) so maybe you can tell us what to do.
> 
> It may not be easy for me to compile with gcc 2.95 on my machine; I'm not
> sure I have a working gcc 2 installation.  I'll give it a try, though.

I will install GCC 3 on our Sparc box and see whether I can reproduce
the problem.  I will have to get up to speed again about what our
problems were exactly on IA-32 re GCC 3 and I'll try to find similar
fixes for Sparc.  That could take a while, tho.

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

Marius> Just doing

Marius>   $ make clean
Marius>   $ make CFLAGS=-g

Marius> should be enough.  You can override the nonessential compiler flags by
Marius> setting CFLAGS on the make command line.

Indeed.  I did keep the -W flags on the principle of changing as little
as possible.  What I found was that -O0 works and -O1 fails, which should
make it somewhat easier to find the faulty optimization.

Marius> I will install GCC 3 on our Sparc box and see whether I can
Marius> reproduce the problem.  I will have to get up to speed again
Marius> about what our problems were exactly on IA-32 re GCC 3 and
Marius> I'll try to find similar fixes for Sparc.  That could take a
Marius> while, tho.

Please let me know.

I was going to suggest a binary-search approach to finding the
problem, but of course that might not work if the optimization problem
shows up in more than one translation unit.  But what should work is
to compile everything with -O0, then recompiling source files one at a
time with -O1 until the failure reappears.


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Marius Vollmer]

Andrew Koenig <ark@research.att.com> writes:

> I was going to suggest a binary-search approach to finding the
> problem, but of course that might not work if the optimization problem
> shows up in more than one translation unit.  But what should work is
> to compile everything with -O0, then recompiling source files one at a
> time with -O1 until the failure reappears.

Yeah, another approach is to just run the thing in the debugger and
trace the segfault back to where the optimization screwed us.

-- 
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3  331E FAF8 226A D5D4 E405


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

Marius> Yeah, another approach is to just run the thing in the
Marius> debugger and trace the segfault back to where the optimization
Marius> screwed us.

That works if the problem isn't a data structure corrupted by
an optimization in an apparently unrelated place.

Anyway, running with -O0 is fine for the time being, so there's
no rush.


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile

Re: guile 1.6.3 documentation installation problem, Solaris 8, gcc 3.2.1 {PS}

[Andrew Koenig]

Marius> Could you try compiling without "-O2"?  Also, if convenient,
Marius> could you compile with GCC 2.95.x?

It appears to work without -O2.  Yecch.  I'm trying -O1 now.

If you know what modules were affected by the IA32 optimization
problems, perhaps the next step would be to see whether those
same modules cause problems on Sparc...


_______________________________________________
Bug-guile mailing list
Bug-guile@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-guile