7 Feb
2025
7 Feb
'25
10:13 a.m.
On Thu Feb 6 17:47:03 2025 +0000, Elizabeth Figura wrote:
That's not really legal unless you make assumptions about the fd cache. Generally speaking, closing a handle may also close its cached fd. This *would* work since we don't mark the inproc sync handles cacheable, but there's not actually any reason they can't be cacheable, and making that assumption feels uncomfortable.
It feels weird to have two caches caching the same things. We should IMO decide that these fds are never cached, or decide that they might be cached, then you could dup the fd if necessary and still close the handle, it'll save you space (this seems to be an important aspect according to the handle field comment) and code.