In Singular spkg-install, disable -pipe on SunOS

[jdemeyer]

The Sun assembler chokes when large files are sent to it via a pipe. Singular by default builds with the -pipe option causing trouble.

In the spkg below, I also add a fix for #12311.

spkg:  http://boxen.math.washington.edu/home/jdemeyer/spkg/singular-3-1-3-3.p5.spkg

[jdemeyer]

Diff for the singular spkg, for review only

[jhpalmieri]

The change here makes sense, although I don't remember seeing a problem with singular on Solaris before. This builds on one OpenSolaris machine, and I'm currently building it on a (slow) Solaris machine. It also works on OS X (as tested in #12369).

[jdemeyer]

Replying to jhpalmieri:

The change here makes sense, although I don't remember seeing a problem with singular on Solaris before.

Probably because you don't use the Sun assembler.

This builds on one OpenSolaris machine

using the GNU assembler.

[jhpalmieri]

I didn't know we even supported building Sage on Solaris using gcc and the Sun assembler.

[jdemeyer]

Replying to jhpalmieri:

I didn't know we even supported building Sage on Solaris using gcc and the Sun assembler.

I didn't know either, but I tried it and this ticket is the only issue I encountered.

[jhpalmieri]

Dave, can you take a look at this?

Jeroen, in case Dave doesn't have time, do you have suggestions about how to test this? On the skynet machine mark, I can try setting the path so that it uses

export PATH=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/bin:$PATH

export LD_LIBRARY_PATH=/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib:/usr/local/gcc-4.5.1/sparc-SunOS-ultrasparc3-sun-as-ld/lib/sparcv9:/usr/local/flex-2.5.35/sparc-SunOS-ultrasparc3-gcc-4.4.3/lib

with corresponding settings for SAGE_FORTRAN and SAGE_FORTRAN_LIB. Will that use the Sun assembler? (By the way, with the default PATH etc. on that machine, I have troubles building libpng...)

[jdemeyer]

Replying to jhpalmieri:

Will that use the Sun assembler?

No, the assembler/linker executables are hardcoded when GCC is compiled. You can see this by:

jdemeyer@mark:~$ gcc -print-prog-name=as
/usr/local/binutils-2.20.1/sparc-SunOS-ultrasparc3-gcc-4.4.3/bin/as

This means that gcc will always use that assembler, which is the GNU assembler:

jdemeyer@mark:~$ /usr/local/binutils-2.20.1/sparc-SunOS-ultrasparc3-gcc-4.4.3/bin/as --version
GNU assembler (GNU Binutils) 2.20.1.20100303
Copyright 2009 Free Software Foundation, Inc.
This program is free software; you may redistribute it under the terms of
the GNU General Public License version 3 or later.
This program has absolutely no warranty.
This assembler was configured for a target of `sparc-sun-solaris2.10'.

I encountered the problem on this ticket while testing #12369. The default as in the PATH is

jdemeyer@mark:~$ which as
/usr/ccs/bin/as
jdemeyer@mark:~$ as -V
as: Sun Compiler Common 10 Patch 05/06/2005

So, when compiling GCC, that assembler will be used by default. This can be overridden using the AS environment variable (similarly LD for the linker).

[jhpalmieri]

Okay, I'll set SAGE_INSTALL_GCC to "yes" on a few machines and build using the tar ball from #12369, and I'll also try to install the old Singular spkg in the resulting Sage installation.

[jhpalmieri]

On hawk (OpenSolaris on x86), after installing beta7 from #12369 with SAGE_INSTALL_GCC=yes, and while running sage -sh:

$ gcc -print-prog-name=as
/usr/ccs/bin/as

Then singular-3-1-3-3.p4 fails to build, while singular-3-1-3-3.p5 succeeds.

I'm also trying on mark, and things there are a little strange: on one hand, the Sage built took only two hours, which is much faster than usual -- an earlier build took 12 hours. Either they upgraded the machine, or the standard gcc build on mark is badly configured, or something else is strange. Tests seem to be going fine (with the Singular spkg from this ticket). On the other hand, if I just try to run Sage, I see this:

$ ./sage
----------------------------------------------------------------------
| Sage Version 5.0.beta7-gcc, Release Date: 2012-03-05               |
| Type notebook() for the GUI, and license() for information.        |
----------------------------------------------------------------------
**********************************************************************
*                                                                    *
* Warning: this is a prerelease version, and it may be unstable.     *
*                                                                    *
**********************************************************************
/home/palmieri/mark/sage-5.0.beta7-gcc/local/lib/python/site.py:158: Warning: 'with' will become a reserved keyword in Python 2.6
'import site' failed; use -v for traceback
/home/palmieri/mark/sage-5.0.beta7-gcc/local/lib/python/linecache.py:127: Warning: 'with' will become a reserved keyword in Python 2.6
/home/palmieri/mark/sage-5.0.beta7-gcc/local/bin/sage-location:185: Warning: 'with' will become a reserved keyword in Python 2.6
  File "/home/palmieri/mark/sage-5.0.beta7-gcc/local/bin/sage-location", line 185
    with open(filename) as file:
            ^
SyntaxError: invalid syntax

If I run ./sage -sh and then gcc, I see this:

$ gcc
bash: /home/palmieri/mark/sage-5.0.beta7-gcc/local/bin/gcc: Invalid argument

Not related to this ticket, but obviously something strange is happening on mark.

[jhpalmieri]

The issues with mark seem to have cleared themselves up somehow. My findings:

• With hawk, the spkg here builds while the old one doesn't.

• With mark and mark2 (Solaris on sparc), both seem to build.

So I think this can get a positive review.