Re: Problem destroying class constants

Hello Cristiano, Sunday, October 5, 2003, 7:15:17 PM, you wrote: > "Cristiano Duarte" <cunha17@uol.com.br> escreveu na mensagem > news:20031004233324.87451.qmail@pb1.pair.com... >> "Marcus BöRger" <mail@marcus-boerger.de> escreveu na mensagem >> news:165216787375.20031005001839@marcus-boerger.de... >> > Ye, constants are only supported for user classes and noone is working >> > on it at the moment. >> >> "Moriyoshi Koizumi" <moriyoshi@at.wakwak.com> escreveu na mensagem >> news:20031005081950iuHFnj@at.wakwak.com... >> > I'm not sure why, but maybe Zeev or Andi will fix it, as the relevant > part >> > is told to be so sensitive that I cannot do with it in a way fine with >> > them. >> Hi Marcus & Moriyoshi, >> >> Thanx for the quick answer. >> But can I manually destroy the hashtable before the engine gets to it ? Is >> there any kind of "event" triggered before the cleanup process ? >> Maybe I could create a "Internal Hashtable Factory" that will cleanup the >> hashtable correctly before the engine cleanup process. Is it possible? > IMHO, >> I think that this aproach has almost no impact in the engine... > In fact I added some functions to zend_hash.c that are capable of storing > internal HashTables and destroying them. Then I modified TSRM.c > (tsrm_shutdown) to call the internal HashTables destroy function and it > worked !!! No more memory leaks nor segmentation faults !!! > Is this approach OK ? It's far to early to completley understand all details but your last patch is the correct way to do. But for me it seems the wrong way calling the shutdown functions from tsrm. They are zend functions and should hence be called from zend shutdown mechanismns. -- Best regards, Marcus mailto:mail@marcus-boerger.de