bug#45198: 28.0.50; Sandbox mode

[Philipp Stephani <p.stephani2@gmail.com>]

Am Mo., 14. Dez. 2020 um 15:44 Uhr schrieb Stefan Monnier
<monnier@iro.umontreal.ca>:
> >>     Returns a process object.
> > Depending how much we care about forward compatibility, it might be
> > better to return an opaque sandbox object (which will initially wrap a
> > process object).
>
> We always use process objects to represent file-descriptors, so I can't
> find any good reason why this one should be different or why an
> implementation might find it difficult to expose a process object.

If we return a single pipe process object here, then there would be no
way to access or wait for the actual Emacs process object. That might
be OK, but maybe it's not. Does Emacs guarantee that (while
(accept-process-output PIPE)) for some pipe process reads the entire
process output and finishes in due time (not too long after the actual
process)? Also, how would we report errors from the subprocess? A pipe
process can't really fail, right?

>
> >>     FUNCTION is called with no arguments and it can use `sandbox-read`
> >>     to read the data sent to the process object via `process-send-string`,
> >>     and `sandbox-reply` to send back a reply to the parent process
> >>     (which will receive it via its `process-filter`).
> > That is, sandbox-read and sandbox-reply just read/write stdin/stdout?
>
> While it may use stdin/stdout internally, I can imagine good reasons why
> we'd want to use some other file descriptors.

Yes, it should in many cases not be stdin/stdout. The standard output
and error are polluted with log messages, and stdin should likely be
closed or empty to avoid Emacs trying to read from it.
It should definitely be possible to create a "magic" pipe process
(similar to the "magic network process" created for systemd socket
activation) that wraps a file descriptor pair passed on the
command-line.
OTOH, for the case at hand using stdout/stderr seems right: the error
messages are printed there, and the parent Emacs process can parse
them.
Or do you suggest sending the error messages in a structured format
(e.g. JSON) over the pipe?

>
> > That would certainly work, but (a) it doesn't really have anything to
> > do with sandboxing, so these functions should rather be called
> > stdin-read and stdout-write or similar,
>
> I think "the right thing" would be to represent the parent as a process
> object inside the child.  I proposed dedicated functions only because
> but when it uses stdin/stdout, providing a process object seems awkward
> to implement.

What would be the difference? It would be a pipe process wrapping some
open file descriptor pair in both cases.