In <[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
> I believe there was a problem with constructors being called in
> reverse link order. Also any object file of the dll can potentially
> have a .init section which would then break badly.
I guess configure could test the constructor semantics, sinc
Alexandre Julliard <[EMAIL PROTECTED]> writes:
> The way it normally works is that the _init() function is built from
> the contents of the .init section, so if you add code in .init it's
> supposed to show up in _init().
Ah, right. This doesn't work on OBSD because __init() is fully
defined in c
Alexandre Julliard writes:
> The dll init routine needs to be in the .init section in order to be
> called first,
Isn't that what attribute((constructor)) does? It places a call to the
routine in the .so's initialization section, whatever that section is,
so that the routine gets invoked during