Thread: HP-UX 11.22 (IA64) IPF Patch for PostgreSQL 7.3.2

Hi folks,

  The first attached file is the HP ANSI C and gcc patch file for
  PostgreSQL 7.3.2 on Itanium Processer Family (IPF) system with HP-UX
  11.22 (a.k.a. 11i version 1.6).

    - Use the inline assembly code in src/include/storage/s_lock.h
    - Add horology test into "expected failures"

  It may be required to modify somewhere in code for gcc due to the
  'create_functions_1' and 'trigers' failures during the regression
  test (refer to regression.diffs). So the attached modifications may
  be experimental for gcc :-)


In case of HP ANSI C Compiler:
------------------------------

  * Must specify the option '+legacy_cpp' in the environment variable
    CPP on the configure command line to use the C preprocessor for
    non-C source:

      % ./configure CC=cc CPP="cc -E +legacy_cpp"

    Otherwise the C preprocessor called in
    src/backend/catalog/genbki.sh doesn't generate the appropriate
    postgres.bki which can be used during the regression test.

    Note that you need at least the C compiler version A.05.38 to use
    the option +legacy_cpp. PHSS_27659 provides the latest version
    A.05.41 as of today.

  * In the regression test, 'float8' and 'geometry' fails. But that
    seems to be acceptable.

In case of GNU gcc:
-------------------

  * I used gcc-core-20030331 (snapshot) on HP-UX 11.22. In the
    regression test, 'float8', 'geometry', 'create_function_1' and
    'triggers' fails. Then I got the attached regression.diffs file.

  * Although gcc 3.2 pre-release version (http://www.hp.com/go/gcc)
    for IPF can compile PostgreSQL 7.3.2 on HP-UX 11.22, the
    regression test fails due to the error "pg_regress: could not drop
    user accounts" after creating database "regression".

Shinji Teragaito

Shinji Teragaito <shinji@kobe.hp.com> writes:
>  The first attached file is the HP ANSI C and gcc patch file for
>  PostgreSQL 7.3.2 on Itanium Processer Family (IPF) system with HP-UX
>  11.22 (a.k.a. 11i version 1.6).

You do realize that this set of diffs shows that the alleged port
is still extremely broken?

Low-order geometry diffs are one thing, but you've got some other
issues to fix there ... I'm not eager to apply any of this until
the regression results look saner.
        regards, tom lane

Shinji Teragaito writes:

>   * Must specify the option '+legacy_cpp' in the environment variable
>     CPP on the configure command line to use the C preprocessor for
>     non-C source:

Please give symptoms of the problem.

Also, regarding this piece of the patch:
 horology/hppa=horology-no-DST-before-1970
+ horology/ia64.*=horology-no-DST-before-1970

Both of these lines are bogus.  They should read horology/.*-hpux=...,
since the horology test outcomes are determined by the operating system,
not the processor.

-- 
Peter Eisentraut   peter_e@gmx.net

>> On Tue, 8 Apr 2003 14:21:30 +0200 (CEST), Peter Eisentraut <peter_e@gmx.net> said:

> Shinji Teragaito writes:
>> * Must specify the option '+legacy_cpp' in the environment variable
>> CPP on the configure command line to use the C preprocessor for
>> non-C source:

> Please give symptoms of the problem.
 http://archives.postgresql.org/pgsql-general/2003-01/msg00842.php
http://archives.postgresql.org/pgsql-general/2003-01/msg00858.php

> Also, regarding this piece of the patch:

>   horology/hppa=horology-no-DST-before-1970
> + horology/ia64.*=horology-no-DST-before-1970

> Both of these lines are bogus.  They should read horology/.*-hpux=...,
> since the horology test outcomes are determined by the operating system,
> not the processor.
 Okay. Thank you.

Best Regards,

Shinji