Make all longjmp and pthread_exit related behavior TBD
#31
Conversation
Typo Co-authored-by: bjorn3 <[email protected]>
|
|
||
| Conversely, if, due to a bug, `longjmp` were called unconditionally, then this | ||
| code would have undefined behavior on all platforms when `c` is false, because | ||
| `foo` would not be a POF. |
There was a problem hiding this comment.
I would expect that the compiler would warn in this scenario, also
There was a problem hiding this comment.
Maybe I'm misunderstanding, but how would the compiler provide a warning for this? The extern function is in C, so the compiler can't determine under what conditions it will longjmp. Or do you mean it would warn in any scenario where a frame is not guaranteed to be a POF at the time it calls a "cancelable" function?
There was a problem hiding this comment.
do you mean it would warn in any scenario where a frame is not guaranteed to be a POF at the time it calls a "cancelable" function?
This is what I meant. I think it's going to be quite rare that you have a function that sometimes cancels but you happen to know that on this particular call it will not and therefore you want to take advantage of that by having a pending destructor. In such a case, adding an #[allow] annotation seems reasonable to me.
There was a problem hiding this comment.
Adding a comment to that effect
Per #30, the initial
"C unwind"RFC will not specify whenlongjmpandpthread_exitis well-defined.