Re: [Bug-zile] Regression tests on stable?

[Gary V. Vaughan]

Hi Reuben,

On 1 Feb 2012, at 20:36, Reuben Thomas wrote:
> On 31 January 2012 14:44, Gary V. Vaughan <address@hidden> wrote:
>> On 31 Jan 2012, at 20:44, Reuben Thomas wrote:
>> 
>>> In any case, it's worth keeping it
>>> alive for now, or at least in touch (for example, other people have
>>> indicated they are interested in hacking it), so it might as well at
>>> least carry tests so it's obvious where it's up to. Currently it
>>> passes all tests and I'm still fixing bugs.
>> 
>> Then keeping the tests in sync too seems like the right approach to me.
> 
> I've synched the tests. This does imply that the one currently failing
> test should be fixed on the C branch too, so I'll watch closely when
> you commit the fix to the Lua branch.

It's not pretty I'm afraid, since it relies on my having turned off curses
keypad mode, installing our own cursor key mode change signal handlers,
and then using tgetstr to populate a lookup table of ESC codes on our own
without curses.  That allows us to always push back onto keybuf the key-by-
key escape sequences so that there is now a single code path for fetching
and interpreting keypresses (previously, handling keys from the buffer was
different to handling keys from curses, which I think has been the root of
this bug all along).

The most relevant patch is 802a38f5 on the zi branch, which I'm in the
process of porting to the lua branch right now :)  Although the port
probably won't land for another day or three...

Cheers,
-- 
Gary V. Vaughan (gary AT gnu DOT org)