Re: cvs: php4 /ext/xml xml.c

"Rob Richards" <rrichards@ctindustries.net> wrote: > From: Moriyoshi Koizumi > > > Indeed. But as long as libxml allocators are designed to be initialised > > only in the xml extension, the order could be a problem in which those > > extensions are initialised on startup. Maybe I can make clones of the > > xmlMemSetup() code and paste them to the PHP_MINIT_FUNCTION() of every > > extension which uses libxml, but it hardly looks like a better solution. > > I'd asked you for more a practical option :) > > We had originally looked at implementing that xml_parser_inited stuff, but > the way it was implemented was basically duplicating the same initialization > code everywhere. I wasn't aware of being the third person here.. > What would be nice would to have a centralized function (it would have to > live outside all the libxml based extensions that every extension could call > on MINIT. It would be re-entrant safe to avoid corruption of globals. Then > the extension could just call the function during its MINIT (as this would > also make it extremely easy to change initialization routines to correspond > with any changes in libxml or add any initialization that would need to take > place). > > This probably does have some draw backs, but it would be ideal. It looks like we(?) need a functionality in TSRM similar to pthread_once(), with which we could make sure that the initialisation routine with an unique integer ID is called only once in the lifetime of the process. As for Win32, I think it's possible to emulate it with TLS stuff though the API doesn't provide a specific function dedicated to that functionality. Moriyoshi