Re: [PHP4BETA] cvs: /php4/ext/odbc config.h.stub config.m4 php3 _velocis.h php_odbc.c php_odbc.h velocis.c

Okay, time to refresh some memories. :-) Back when 3.0 was in early alpha (or pre-alpha, I don't remember), we used these HAVE_XXX symbols in the internal functions array, to determine whether or not the functions from that extension were to be added to the global function table. AFAIR, that was _the_ reason we switched from the de-facto standard of "undef/define 1" to "define 0/define 1", because we were actually using the values. But after this was changed to defining module_entry_ptr to NULL instead, we didn't change it back. This has confused a number of people into thinking that "define 0/define 1" is the way to do it. They get bitten as soon as they use/cook/look at code from projects following the de-facto standard. This is at least my "historical perception" of this. Maybe we should do a complete cleanup of this after b3? We could rename any HAVE_XXX symbols that have the values 0 or 1 to either "WITH_XXX" or "ENABLE_XXX", and leave the HAVE_XXX "namespace" to undef/define 1 symbols. If we document this in a visible place, that would hopefully be the end of the confusion. - Stig On Wed, 10 Nov 1999, Rasmus Lerdorf wrote: | On Tue, 9 Nov 1999, Stig Bakken wrote: | > On Tue, 9 Nov 1999, Rasmus Lerdorf wrote: | > | > | > These changes were done for consistency. I'll update config.w32.h. | > | | > | Then you also need to change every single define in php_config.h | > | > I changed ext/odbc/config.h.stub, from which php_config.h is built, so the | > HAVE_NNN symbols used in ext/odbc default to undefined. Or do you mean | > something else? | | Ok, perhaps I am being completely daft here, but all the HAVE_* defines in | php_config.h define to either 0 or 1, do they not? ie. a partial | copy-paste from it: | | /* Define if you want the LDAP directory interface */ | #define HAVE_LDAP 0 | /* Define if you want the MCAL extension */ | #define HAVE_MCAL 0 | /* define if you want to use the mcrypt extension */ | /* #undef HAVE_LIBMCRYPT */ | /* define if you want to use the mhash extension */ | /* #undef HAVE_LIBMHASH */ | #define HAVE_MSQL 0 | #define MSQL1 0 | #ifndef HAVE_MYSQL | #define HAVE_MYSQL 0 | #endif | /* define if you want to use the oci8 extension */ | /* #undef HAVE_LIBOCI8 */ | #define HAVE_OCI8 0 | | #define HAVE_SOLID 0 | #define HAVE_IBMDB2 0 | #define HAVE_IODBC 0 | #define HAVE_ESOOB 0 | #define HAVE_UNIXODBC 0 | #define HAVE_OPENLINK 0 | #define HAVE_ADABAS 0 | #define HAVE_CODBC 0 | #define HAVE_UODBC 0 | #define HAVE_EMPRESS 0 | #define HAVE_VELOCIS 0 | /* Define if you have the Oracle database client libraries */ | #define HAVE_ORACLE 0 | /* define if you want to use the PCRE extension */ | #define HAVE_PCRE 0 | | | That's what I meant when I said you'd need to change all the other ones as | well. We have always used this approach, and if you are going to change | the ODBC case to #define blah 1/#undef blah then we should be consistent | and do it throughout. Personally I don't see why what we have isn't | sufficient. | | -Rasmus |