Re: [9front] [PATCH] rio: add -label wctl param; improve window(1)

[kvik@a-b.xyz]

I appreciate you taking the time to review.

> otherwise you got a classical format string attack.

Ouch!

> about the -label flag in wctl message. i'm not so sure it is a good
> idea. adding new flags there means a breaking change and updating rio
> needs to be synchronized.

> i think running window recursively is not really the issue.

It's not an issue per se, but it seems we are only doing it the
roundabout way because of wanting to label the new window while missing
the -label parameter to do so.  We also do the environment cleanup, but
this only affects the -m case so it can be moved there.

I'd argue that the label is a window property just like the others we
are already setting through the 'new' wctl.  I see no reason why it
should be handled specially from the rest, or alternatively why other
parameters shouldn't be set in a similar fashion by writing to /dev/wctl
in new window's namespace.

Introducing breakage is quite unfortunate, though I'm not entirely
convinced it should be a single show stopper.

> also we have the problem with quoting. (-cd also suffers from this).

Yes, the field parser is janky and could be helped with a rewrite in
terms of getfields(2).  I can do it independently of this other stuff.

> the real issue is that we have to properly preserve the arguments. 

> the bug is when we do $"cmd which destroys list tokenization.

> a solution can be accomplished with rc's whatis, which lets rc
> serialize a variable as a string that we can tunnel thru the wctl
> text interface.

Oh, wow, whatis is gold and I somehow missed the list serialization
feature completely!  I'll make use of it in a followup patch.