Re: __builtin_va_list

From: Bruce A. Mallett <bam_at_NightStorm.com>
Date: 1 Apr 2002 21:56:29 -0800
Message-ID: <1ad12056.0204012156.3f8ac821@posting.google.com>

Believe it or not, I got it to work. Here is how ...

First I created a file named /usr/local/include/stdarg.h having the following contents:

• CUT CUT ==================== #if !defined( _H_STDARG_ ) #define _H_STDARG_ #define __builtin_va_list void ** #include "/usr/lib/gcc-lib/i586-mandrake-linux-gnu/2.96/include/stdarg.h" #endif /* _H_STDARG_ */
• CUT CUT ====================

You will obviously need to substitute the full path to the *real* stdarg.h for the define of "__builtin_va_list" in the above file (you might need to use "find / -name stdarg.c 2>/dev/null" to locate it).

I then put "/usr/local/include" at the *head* of my sys_include list. Thus my pcscfg.cfg file looks like this:

• CUT CUT ==================== sys_include=(/usr/local/include,$ORACLE_HOME/precomp/public,/usr/lib/gcc-lib/i586-mandrake-linux-gnu/2.96/include,/usr/include,/usr/include/sys) include=$ORACLE_HOME/rdbms/demo include=$ORACLE_HOME/rdbms/public include=$ORACLE_HOME/precomp/public include=$ORACLE_HOME/precomp/syshdr CODE=ANSI_C
• CUT CUT ====================

The idea here is to trick Pro*C into finding my stdarg.h which gives it something reasonable to chew on in terms of a definition for "__builtin_va_list". When gcc runs it of course goes directly to the real stdarg.h.

Hope that works for you too!

• Bruce bam_at_DELETEnightstormTHIS.com

       (remove UPPERcase letters to get a real address)

mngarimu_at_hotmail.com (Michael Ngarimu) wrote in message news:<a4ef9fb1.0203081504.48c982_at_posting.google.com>...
> Okay, I've seen a hand full of posts/web pages dated around 2000-2001
> relating to the Oracle Pro*C/C++ precompiler (at least on version
> 8.1.7 anyway) barfing on __builtin_va_list... The only suggestion
> anyone has had is to use PARSE=NONE (or possibly PARSE=PARTIAL).
>
> Unfortunately, the company I work for has _a_lot_ of existing Oracle
> Pro*C code that requires PARSE=FULL (due to using function formal
> arguments in EXEC SQL statements).
>
> My problems exists with the following software version:
> - Oracle 8.1.7.1 for Linux
> - RedHat 7.2
> - glibc 2.1 (the compat-egcs compat-glibc etc etc patches have been
> applied)
> - glibc 2.2 (I tried it with 2.2 because the compat patch mentioned
> that you could still compile and link with newer versions of glibc -
> you only needed to link with the compatibility version for the actual
> Oracle binaries etc - or did I get that wrong)
>
> Does anyone have a solution for this yet?
>
> Mike.
Received on Mon Apr 01 2002 - 23:56:29 CST