Re: [cxx-abi-dev] One-time Construction API (3.3.2)

[Dennis Handly <dhandly@xxxxxxxxxx>]

>From: Jason Merrill <jason@xxxxxxxxxx>
>primarily because if construction of the object throws, we need to call
>__cxa_guard_abort, but if destruction of any temporaries throws, we need to
>call __cxa_guard_release.

Hmm, we seem to do the same for both.  And it is left on the atexit list
for the latter.

>  To make this work, we need to do something like
>if (__cxa_guard_acquire (&guard)) {

You didn't optimize by checking the guard before calling
_cxa_guard_acquire.

>The flag is necessary to avoid non-nested overlapping EH regions.

We don't use a flag.  We use try/catch:
    try {
        /* initialize */
        __cxa_atexit(dtor);
        __cxa_guard_release(&guard);
    }
    catch(...) {
          __cxa_guard_abort(&guard);
    }

>2.8 refers to "the first byte", while 3.3.2 refers to "the low-order
>byte".  These are the same on little-endian targets, but not big-endian.

Are you sure?  It says byte with lowest address.  Not low-order byte.
First and lowest are the same for big-endian.

>I've been thinking about how to implement the actual locking, and the best
>I've come up with is to have the guard be an index into an array of mutexes

We use one mutex for all statics.  The same one that synchronizes the
__cxa_atexit list.

>with magic values for uninitialized and "creating
>mutex"; in the latter case another thread that comes to it should just spin
>until it takes on another value.

We do use the second byte to handle (stop) recursion.  Though this seems
flawed in that it won't wait until constructed.  :-(

>Has anyone actually implemented this?
Jason

Naturally, 3 years ago.