cleanup handlers and longjmp

[Martin v. Loewis]

> That's not to say that it would not be nice to have longjmp, cleanup
> handlers and C++ exception handling all play together nicely, for
> writing programs that are specific to GNU/Linux.

Currently, Posix is silent on the issue of C++ in general, and
exceptions in particular. Likewise, C++ is silent on the issue of
threads.

It is likely that vendors that offer to combine both also attempt to
offer "proper" unwinding of C++ exceptions in a thread, in the long
run. So I expect that future revisions of the relevant standards are
more specific on the issue of exceptions and threads.

> No, some extensions are based on syntax errors and diagnosable
> constraint violations. E.g. gcc's support for zero length arrays,
> nested functions, or labels without a statement.

Oops, you're right.

Regards,
Martin