This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: static constructors/destructors and atexit


                                                         This was
    discussed in the thread `p2736.C Is dtor order guaranteed?' in this
    mailing list

Thanks for the pointer.  It looks like this ended with Mike Stump's
comment

    for C++, we would need to extend any atexit with just 64 entries.
    This is like support for identifiers with only 31 characters in them.

so the idea is use gcc's atexit on systems that implement limited atexits?

I'm not sure how easy that is going to be; I'm afraid some places it's
going to be quite hard.  Not that I have a better idea.

In our case, not interleaving atexit is going to be less painful than
trying to get a correct implementation.  On systems where atexit is
limited, the current behavior is to just not call the destructors; that's
fatal for us.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]