Not OK: perl 5.00559 on IP32-irix 6.5 (UNINSTALLED)

[p5pRT]

Migrated from rt.perl.org#1142 (status was 'resolved')

Searchable as RT1142$

[p5pRT]

From kstar@o2.chapin.edu

---

  `make all' fails when building Socket. This platform usually
works well. Details to follow. Pointers welcomed.

  Making Socket (dynamic)
make[1]​: Entering directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
make[1]​: Leaving directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
make[1]​: Entering directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
gcc -c -D_BSD_TYPES -D_BSD_TIME -D_POSIX_C_SOURCE -I/usr/local/include -DLANGUAGE_C -O3 -DVERSION=\"1.71\" -DXS_VERSION=\"1.71\" -fpic -I../.. Socket.c
In file included from Socket.xs​:32​:
/usr/include/netinet/tcp.h​:52​: duplicate member `th_off'
/usr/include/netinet/tcp.h​:53​: duplicate member `th_x2'
make[1]​: *** [Socket.o] Error 1
make[1]​: Leaving directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
make​: *** [lib/auto/Socket/Socket.so] Error 2

---

---

Site configuration information for perl 5.00559​:

Configured by kstar at Mon Aug 2 14​:49​:12 EDT 1999.

Summary of my perl5 (revision 5.0 version 5 subversion 59) configuration​:
  Platform​:
  osname=irix, osvers=6.5, archname=IP32-irix
  uname='irix o2 6.5 11051733 ip32 '
  config_args='-des -Dcf_email=kstar@​chapin.edu -Dcc=gcc'
  hint=recommended, useposix=true, d_sigaction=define
  usethreads=undef useperlio=undef d_sfio=undef
  use64bits=undef usemultiplicity=undef
  Compiler​:
  cc='gcc', optimize='-O3', gccversion=2.8.1
  cppflags='-D_BSD_TYPES -D_BSD_TIME -D_POSIX_C_SOURCE -I/usr/local/include -DLANGUAGE_C'
  ccflags ='-D_BSD_TYPES -D_BSD_TIME -D_POSIX_C_SOURCE -I/usr/local/include -DLANGUAGE_C'
  stdchar='unsigned char', d_stdstdio=define, usevfork=false
  intsize=4, longsize=4, ptrsize=4, doublesize=8
  d_longlong=define, longlongsize=8, d_longdbl=define, longdblsize=8
  alignbytes=8, usemymalloc=y, prototype=define
  Linker and Libraries​:
  ld='gcc', ldflags =' -L/usr/local/lib -L/usr/gnu/lib'
  libpth=/usr/local/lib /usr/gnu/lib /usr/lib /lib
  libs=-lgdbm -ldb -lm -lrt -lc
  libc=, so=so, useshrplib=false, libperl=libperl.a
  Dynamic Linking​:
  dlsrc=dl_dlopen.xs, dlext=so, d_dlsymun=undef, ccdlflags=' '
  cccdlflags='-fpic', lddlflags='-shared -L/usr/local/lib -L/usr/gnu/lib'

Locally applied patches​:
 

---

@​INC for perl 5.00559​:
  lib
  /usr/local/lib/perl5/5.00559/IP32-irix
  /usr/local/lib/perl5/5.00559
  /usr/local/lib/site_perl/5.00559/IP32-irix
  /usr/local/lib/site_perl
  .

---

Environment for perl 5.00559​:
  HOME=/usr/people/kstar
  LANG=en_US
  LANGUAGE (unset)
  LD_LIBRARY_PATH (unset)
  LOGDIR (unset)
  PATH=​:/usr/people/kstar/bin​:/usr/local/bin​:/usr/gnu/bin​:/usr/local/bin​:/usr/sbin​:/usr/bsd​:/sbin​:/usr/bin​:/usr/bin/X11​::/usr/etc
  PERL_BADLANG (unset)
  SHELL=/bin/sh

[p5pRT]

From @jhi

Kurt D. Starsinic writes​:

This is a build failure report for perl from kstar@​chapin.edu,
generated with the help of perlbug 1.26 running under perl 5.00559.

-----------------------------------------------------------------

\`make all' fails when building Socket\.  This platform usually

works well. Details to follow. Pointers welcomed.

    Making Socket \(dynamic\)

make[1]​: Entering directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
make[1]​: Leaving directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
make[1]​: Entering directory `/usr/people/kstar/Projects/perl5.005_59/ext/Socket'
gcc -c -D_BSD_TYPES -D_BSD_TIME -D_POSIX_C_SOURCE -I/usr/local/include -DLANGUAGE_C -O3 -DVERSION=\"1.71\" -DXS_VERSION=\"1.71\" -fpic -I../.. Socket.c
In file included from Socket.xs​:32​:
/usr/include/netinet/tcp.h​:52​: duplicate member `th_off'
/usr/include/netinet/tcp.h​:53​: duplicate member `th_x2'

Hmph. If you could figure out why SGI has a broken <netinet/tcp.h>
and how could we circumcode it, that would be nice.

--
$jhi++; # http​://www.iki.fi/jhi/
  # There is this special biologist word we use for 'stable'.
  # It is 'dead'. -- Jack Cohen

[p5pRT]

From @jhi

: In file included from Socket.xs​:32​:
: /usr/include/netinet/tcp.h​:52​: duplicate member `th_off'
: /usr/include/netinet/tcp.h​:53​: duplicate member `th_x2'
: make[1]​: *** [Socket.o] Error 1

I am looking at IRIX 6.5's <netinet/tcp.h> (haven't yet tried
compiling _60) and I'm rather baffled by the above.

[snip]

47 #if BYTE_ORDER == LITTLE_ENDIAN
48 u_char th_x2​:4, /* (unused) */
49 th_off​:4; /* data offset */
50 #endif
51 #if BYTE_ORDER == BIG_ENDIAN
52 u_char th_off​:4, /* data offset */
53 th_x2​:4; /* (unused) */
54 #endif

[snap]

I didn't know MIPS has, ummm, circular registers? :-)

--
$jhi++; # http​://www.iki.fi/jhi/
  # There is this special biologist word we use for 'stable'.
  # It is 'dead'. -- Jack Cohen

[p5pRT]

From [Unknown Contact. See original ticket]

Jarkko Hietaniemi <jhi@​iki.fi> writes​:

I am looking at IRIX 6.5's <netinet/tcp.h> (haven't yet tried
compiling _60) and I'm rather baffled by the above.

[snip]

47 #if BYTE_ORDER == LITTLE_ENDIAN
48 u_char th_x2​:4, /* (unused) */
49 th_off​:4; /* data offset */
50 #endif
51 #if BYTE_ORDER == BIG_ENDIAN
52 u_char th_off​:4, /* data offset */
53 th_x2​:4; /* (unused) */
54 #endif

Wild guess as to why you get dups - none of those symbols are defined
as you have not included <endian.h> or whatever - so 0 == 0 in both cases.

[snap]

I didn't know MIPS has, ummm, circular registers? :-)

This is a real pain that I think ANSI C has foisted on us hardware types.
When your C compiler is "big endian" it is _supposed_ to start allocating
fields at MS end of word. When "little endian" then it is supposed to start
from LS end.

However those of use that design chips don't change bit positions based
on endian mode chip is in (too many wires and gates that people don't
want to pay for).

So to have a struct with bitfields describe a hardware register with (say)
th_off in MS nibble and th_x2 in LS nibble you get messes like the above.

The same kind of mess can occur when describing network packets which have
a defined order of fields "on-the-wire".

--
Nick Ing-Simmons

[p5pRT]

From @jhi

Wild guess as to why you get dups - none of those symbols are defined
as you have not included <endian.h> or whatever - so 0 == 0 in both cases.

<slap> Ahh, that might be it. Kurt, could you verify? Add #include
<sys/endian.h> before #include <netinet/tcp.h> in Socket.xs and retry.

--
$jhi++; # http​://www.iki.fi/jhi/
  # There is this special biologist word we use for 'stable'.
  # It is 'dead'. -- Jack Cohen

[p5pRT]

From [Unknown Contact. See original ticket]

On Wed, Aug 04, 1999 at 12​:06​:15AM +0300, Jarkko Hietaniemi wrote​:

Wild guess as to why you get dups - none of those symbols are defined
as you have not included <endian.h> or whatever - so 0 == 0 in both cases.

<slap> Ahh, that might be it. Kurt, could you verify? Add #include
<sys/endian.h> before #include <netinet/tcp.h> in Socket.xs and retry.

  That doesn't do it. I'm looking into it. Watch this space.

  Peace,
* Kurt Starsinic (Kurt.Starsinic@​isinet.com) --------- Technical Specialist *
| `. . . we are clearly living in a period of historical discontinuity |
| in which the structures of the past seem to have lost their power |
| to determine the future.' -- Pekka Tarjanne, ITU Director-General |
Institute for Scientific Information http​://www.isinet.com/