Problems with 8.4, FLOAT8PASSBYVAL and x86_64 GNU/Linux - Mailing list pgsql-general

From Rafael Martinez
Subject Problems with 8.4, FLOAT8PASSBYVAL and x86_64 GNU/Linux
Date
Msg-id 4A5EEF88.2010807@usit.uio.no
Whole thread Raw
Responses Re: Problems with 8.4, FLOAT8PASSBYVAL and x86_64 GNU/Linux  (Peter Eisentraut <peter_e@gmx.net>)
List pgsql-general
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello

We have a C function that works without problems when we use it with
postgresql 8.3 (32/64bit) and with postgreSQL 8.4 (32bit).

But with 8.4 in a 64bit server, postgresql generates this error when we
try to install the function:

ERROR:  incompatible library "/usr/local/lib/pg_uname_8.4.so": magic
block mismatch
DETAIL:  Server has FLOAT8PASSBYVAL = true, library has false.

This is the query we use to install the function:

CREATE OR REPLACE FUNCTION pg_uname(text) RETURNS text
    AS '/usr/local/lib/pg_uname_8.4.so', 'pg_uname'
    LANGUAGE c STRICT;

The only reference we have found about this is in the 8.4 changelog:

" ... Pass float8, int8, and related datatypes by value inside the
server on 64-bit platforms (Zoltan Boszormenyi)

Add configure option --disable-float8-byval to use the old behavior. As
above, this change might break old-style external C functions. ..."

We cannot find more information in the documentation about things we
should change so the function works with 8.4 without using
- --disable-float8-byval under the compilation of postgresql.

What do we need to change in the function to avoid this problem?

The function we are talking about is this one:
- -------------------------------------------------------------------
#include "postgres.h"
#include <string.h>
#include "fmgr.h"
#include <sys/utsname.h>

#ifdef PG_MODULE_MAGIC
PG_MODULE_MAGIC;
#endif

PG_FUNCTION_INFO_V1(pg_uname);

Datum
pg_uname(PG_FUNCTION_ARGS)
{
    text *argument = PG_GETARG_TEXT_P(0);
    size_t argumentlen = VARSIZE(argument)-VARHDRSZ;

    text *result = (text *) palloc(256);
    char *option = (char *) palloc(argumentlen+1);

    char sysname[] = "sysname";
    char nodename[] = "nodename";
    char release[] = "release";
    char version[] = "version";
    char machine[] = "machine";
    char null[] = "null";

    struct utsname uname_pointer;
    uname(&uname_pointer);

    memcpy(option,VARDATA(argument),argumentlen);
    option[argumentlen] = '\0';

    if (strcmp(option,sysname) == 0){
      SET_VARSIZE(result, strlen(uname_pointer.sysname) + VARHDRSZ);

memcpy(VARDATA(result),uname_pointer.sysname,strlen(uname_pointer.sysname));
    }
    else if (strcmp(option,nodename) == 0){
      SET_VARSIZE(result, strlen(uname_pointer.nodename) + VARHDRSZ);

memcpy(VARDATA(result),uname_pointer.nodename,strlen(uname_pointer.nodename));
    }
    else if (strcmp(option,release) == 0){
       SET_VARSIZE(result, strlen(uname_pointer.release) + VARHDRSZ);

memcpy(VARDATA(result),uname_pointer.release,strlen(uname_pointer.release));
    }
    else if (strcmp(option,version) == 0){
      SET_VARSIZE(result, strlen(uname_pointer.version) + VARHDRSZ);

memcpy(VARDATA(result),uname_pointer.version,strlen(uname_pointer.version));
 }
    else if (strcmp(option,machine) == 0){
      SET_VARSIZE(result, strlen(uname_pointer.machine) + VARHDRSZ);

memcpy(VARDATA(result),uname_pointer.machine,strlen(uname_pointer.machine));
    }
    else{
      memcpy(VARDATA(result),null,sizeof(null));
    }

    pfree(option);
    PG_RETURN_TEXT_P(result);
}

- -------------------------------------------------------------------
Thanks in advance.

regards,
- --
 Rafael Martinez, <r.m.guerrero@usit.uio.no>
 Center for Information Technology Services
 University of Oslo, Norway

 PGP Public Key: http://folk.uio.no/rafael/
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.7 (GNU/Linux)

iD8DBQFKXu+GBhuKQurGihQRAqP/AJ4p7DPCwMYzEdO/cykEqf2QpSuiygCeMp9D
GGwluLrii7FRy6+GLo86P3I=
=APUL
-----END PGP SIGNATURE-----

pgsql-general by date:

Previous
From: Marek Lewczuk
Date:
Subject: Getting list of tables used within a query
Next
From: "Albe Laurenz"
Date:
Subject: Re: [PERFORM] Concurrency issue under very heay loads