If you have the interest in discovering more, given that you already have a solution, then you could try installing the latest CPAN version (1.63) and see if that also cures the problem with using threads->self().
I have 1.63 installed and it's not an easy task to back that out, but if that cures the problem for you it's probably safe to conclude it is a old problem that has already been fixed.
Strange that I never encountered it in the couple of years that I've been using 5.8.6/5.8.8, but then I always use threads->self->tid for identification purposes and have never had reason to obtain or reference a thread handle from within it's own thread procedure.
As I mentioned, there is a certain logic to it. If the thread procedure contains a reference to itself, then you have a circular reference, and that is a well-known problem that prevents destructors being called.
I wonder how they cured it? The usual method is to weaken the reference (artificially drop the reference count), but looking inside 1.63 they appear to mortalise before returning it--which I thought meant that the ref count was increased? But then, there is a bunch of the internal shenanigans that is tantamount to black magic as far as my understanding goes.
In reply to Re^7: Hanging Threads
by BrowserUk
in thread Hanging Threads
by WalkingZero
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |