Re: Interrupts in completion, traps in _main_complete

[Peter Stephenson <p.stephenson@samsung.com>]

On Mon, 8 Dec 2014 20:43:10 -0800
Bart Schaefer <schaefer@brasslantern.com> wrote:

> On Dec 8,  8:27pm, Peter Stephenson wrote:
> }
> } I'm probably missing a lot, but it struck me that if you're aborting out
> } of completion completely, as it were, then the stuff that doesn't get
> } executed doesn't typically matter all that much.  The stuff you really
> } want to get right on exit --- options, IFS, etc. --- is already local.
> 
> I'm looking at LISTPROMPT, MENUPROMPT, MENUSCROLL, MENUSELECT, MENUMODE,
> and ZLS_COLORS.  Of which I guess only ZLS_COLORS gets reset right now,
> but on an interrupt even it won't get restored.
> 
> Incidentally, MENUMODE appears to be entirely undocumented.

Apart from ZLE_COLORS it looks like we're not in imminent danger of making
anything worse.

I don't really understand what's going on with ZLE_COLORS, however.  In
the majority of cases it just seems to be saved and restored as if it
were local.  The only exception seems to be if the show-ambiguity style
is set it gets saved for use in another completion.  What is it about
show-ambiguity that requires this?  Is it so that if we redisplay a
listing without going through _main_complete again it appears correct?

I wonder if we're close to the point where applying this branch to the
master is reasonable.  It doesn't have to be 100% right, just good
enough that people who use the bleeding edge don't fall flat on their
faces.

pws