On Fri, Sep 19, 2003 at 12:55:21AM +0100, Philip Blundell wrote: > On Fri, 2003-09-19 at 00:40, Daniel Jacobowitz wrote: > > OK, one is the PLT reference, the other is the global data. Shouldn't > > that global data be fixed up by the loader? > > For that to work, the loader would have to remember that it had seen an > unresolved weak reference to those symbols in the GOT, then go back and > fill in the appropriate entries once it noticed that an object > containing those symbols had been loaded. As far as I know, it doesn't > do that. (If it did, I think you could easily get into sticky > situations when objects were later unloaded.)
Ouch - you're right of course, later it would be unloaded and the global data would still point at it. The linker shouldn't go fixing it up - it might be an initial value of a writable variable, and having dlopen clobber it would be upsetting. -- Daniel Jacobowitz MontaVista Software Debian GNU/Linux Developer -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
