[Python-Dev] pthreads question: typedef ??? pthread_t and hacky return statements

[Tim Peters]

[/F]
> the compiler doesn't warn about bits being lost -- it complained
> because the code was returning a pointer from a function declared
> to return a long integer.
>
> (explicitly casting the pthread_t to a long gets rid of the warning).

For the umpty-umpth time, the code with the simple cast to long is what was
there originally.  The convoluted casting was added later to stop "Alpha
OSF/1" compiler complaints.  Perhaps the compiler no longer complains,
though, or perhaps the one or two people who have tried it since don't have
a version of the compiler that cares about it.

> ...
> frankly, the code is just plain broken.  instead of adding even more dumb
> hacks, just fix it.  here's how it should be done:
>
>     return (long) pthread_self(); /* look! no variables! */

Fine by me, provided that works on all current platforms, and it's
understood that the function is inherently hosed anyway (the cast to long is
inherently unsafe, and we're still doing nothing to meet the promise in the
docs that this function returns a non-zero integer).

> or change
>
>  /* Jump through some hoops for Alpha OSF/1 */
>
> to
>
>  /* Jump through some hoops because Tim Peters wants us to ;-) */

Also fine by me, provided that works on all current platforms, and it's
understood that the function is inherently hosed anyway (the cast to long is
inherently unsafe, and we're still doing nothing to meet the promise in the
docs that this function returns a non-zero integer).