nobu / ruby-lang.org wrote:
> Is "atomically" important?
> If no, it'd be possible to implement without kernel support.

Yes, atomicity is an important distinction.

We should not try to implement this non-atomically without kernel
support; it will quietly break code which expects the atomicity.

I am recalling similar problems around posix_fallocate in glibc:
https://sourceware.org/bugzilla/show_bug.cgi?id=15661

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>