perlirix

CONTENTS

• NAME
• DESCRIPTION
  • Building 32-bit Perl in Irix
  • Building 64-bit Perl in Irix
  • About Compiler Versions of Irix
  • Linker Problems in Irix
  • Malloc in Irix
  • Building with threads in Irix
  • Irix 5.3
• AUTHOR

NAME

perlirix - Perl version 5 on Irix systems

DESCRIPTION

This document describes various features of Irix that will affect how Perl version 5 (hereafter just Perl) is compiled and/or runs.

Building 32-bit Perl in Irix

Use

sh Configure -Dcc='cc -n32'

to compile Perl 32-bit. Don't bother with -n32 unless you have 7.1 or later compilers (use cc -version to check).

(Building 'cc -n32' is the default.)

Building 64-bit Perl in Irix

Use

sh Configure -Dcc='cc -64' -Duse64bitint

This requires require a 64-bit MIPS CPU (R8000, R10000, ...)

You can also use

sh Configure -Dcc='cc -64' -Duse64bitall

but that makes no difference compared with the -Duse64bitint because of the cc -64.

You can also do

sh Configure -Dcc='cc -n32' -Duse64bitint

to use long longs for the 64-bit integer type, in case you don't have a 64-bit CPU.

If you are using gcc, just

sh Configure -Dcc=gcc -Duse64bitint

should be enough, the Configure should automatically probe for the correct 64-bit settings.

About Compiler Versions of Irix

Some Irix cc versions, e.g. 7.3.1.1m (try cc -version) have been known to have issues (coredumps) when compiling perl.c. If you've used -OPT:fast_io=ON and this happens, try removing it. If that fails, or you didn't use that, then try adjusting other optimization options (-LNO, -INLINE, -O3 to -O2, etcetera). The compiler bug has been reported to SGI. (Allen Smith <[email protected]>)

Linker Problems in Irix

If you get complaints about so_locations then search in the file hints/irix_6.sh for "lddflags" and do the suggested adjustments. (David Billinghurst <[email protected]>)

Malloc in Irix

Do not try to use Perl's malloc, this will lead into very mysterious errors (especially with -Duse64bitall).

Building with threads in Irix

Run Configure with -Duseithreads which will configure Perl with the Perl 5.8.0 "interpreter threads", see threads.

For Irix 6.2 with perl threads, you have to have the following patches installed:

1404 Irix 6.2 Posix 1003.1b man pages
1645 Irix 6.2 & 6.3 POSIX header file updates
2000 Irix 6.2 Posix 1003.1b support modules
2254 Pthread library fixes
2401 6.2 all platform kernel rollup

IMPORTANT: Without patch 2401, a kernel bug in Irix 6.2 will cause your machine to panic and crash when running threaded perl. Irix 6.3 and later are okay.

Thanks to Hannu Napari <[email protected]> for the IRIX
pthreads patches information.

Irix 5.3

While running Configure and when building, you are likely to get quite a few of these warnings:

ld:
The shared object /usr/lib/libm.so did not resolve any symbols.
      You may want to remove it from your link line.

Ignore them: in IRIX 5.3 there is no way to quieten ld about this.

During compilation you will see this warning from toke.c:

uopt: Warning: Perl_yylex: this procedure not optimized because it
      exceeds size threshold; to optimize this procedure, use -Olimit
      option with value >= 4252.

Ignore the warning.

In IRIX 5.3 and with Perl 5.8.1 (Perl 5.8.0 didn't compile in IRIX 5.3) the following failures are known.

Failed Test                  Stat Wstat Total Fail  Failed  List of Failed
-----------------------------------------------------------------------
../ext/List/Util/t/shuffle.t    0   139    ??   ??       %  ??
../lib/Math/Trig.t            255 65280    29   12  41.38%  24-29
../lib/sort.t                   0   138   119   72  60.50%  48-119
56 tests and 474 subtests skipped.
Failed 3/811 test scripts, 99.63% okay. 78/75813 subtests failed,
   99.90% okay.

They are suspected to be compiler errors (at least the shuffle.t failure is known from some IRIX 6 setups) and math library errors (the Trig.t failure), but since IRIX 5 is long since end-of-lifed, further fixes for the IRIX are unlikely. If you can get gcc for 5.3, you could try that, too, since gcc in IRIX 6 is a known workaround for at least the shuffle.t and sort.t failures.

AUTHOR

Jarkko Hietaniemi <[email protected]>

Please report any errors, updates, or suggestions to [email protected].