[Toybox] [PATCH] vi: added backspace

[Rob Landley]

On 2/3/23 13:29, Jarno Mäkipää wrote:
> On Fri, Feb 3, 2023 at 6:09 PM Rob Landley <rob at landley.net> wrote:
>>
>> On 2/1/23 15:18, enh via Toybox wrote:
>> > heh, to be clear: i wasn't "dissing" drive-by patching. it accounts
>> > for at least 80% of my entire career :-)
>>
>> https://landley.net/toybox/downloads/binaries/mkroot/0.8.9/linux-patches/
>>
>> > i use the term just to acknowledge that for some things -- like this
>> > -- there isn't anyone else who's actually working on the thing full
>> > time, which is my personal rationale for wanting "the simplest thing
>> > that could possibly work", and why my definition of "simple" is
>> > something like "the most easily understood by an average programmer
>> > who hasn't seen this particular code before".
>>
>> I mean to take custody of this thing this year. I'm trying to get to a 1.0
>> release, which means (among other things) emptying the pending directory entirely.
> 
> Well it might be complete rewrite when you get to it, but hopefully
> its better than starting from nothing :)

Indeed.

>> I do admit vi is something I haven't worked out how to regression test yet, but
>> dogfooding it would presumably cover a multitude of sins. (I do edit both my
>> code and my blog in vi...)
> 
> I was trying to code with toybox vi, but I seem to be too blind and I
> need syntax hilighting. And I really dont wanna implement it here. Its
> not in scope of this.

Syntax highlighting that's built _in_, no. But a syntax highlight thingy based
on regexes sounds like it might make sense?

Although telling it that bash has if/then/fi stages starts to sound sed-ish and
I'd want to pace and stare at trees quite a bit to come up with an actual idea
there. The point would be vi.c remaining simple and the /etc/vi/c.hi and
/etc/vi/bash.hi files it imported would have all the brains...

But THAT can of worms belongs AFTER toysh. And after other cleanup/rewrite of
the basic vi functionality. And I still need an awk for a bunch of build scripts...

>> > (and, yes, in addition to the open() error -- which at least led to a
>> > small simplification of the code -- i've shot myself in the foot by
>> > forgetting that there even are vi tests, not running them, and
>> > breaking them with my recent commit, which i'll have to do something
>> > about before i can sync to AOSP.
>>
>> Huh, I forgot that too. :)
>>
>> Ok, "vi -s" is a good start...
> 
> "vi -s" is ok for text edit operation testing, and tests with weird
> files and such. But visual testing I think using manually with
> different terminal emulators is only way.

I have a vague idea for automated pty master/slave tests, but it's another can
of worms I dowanna open just now, and it's the kind of thing you have to design
as you go. Start with a vague idea and let the implementation guide you or
you're just gonna wind up ripping it all out and starting over repeatedly. (I
mean, that's likely to happen twice _anyway_, but the "I dunno what I'm doing
until I've done it" smells strong with this one...)

> At some point I forked ST to print out escape codes for me when I was
> trying to debug some drawing issues. But fortunately currently drawing
> is just one function (monsterous that needs to be rewritten for sure).
> 
> I think that most work is actually going through all movement commands
> corner cases, things like cw and dw has different movement.
> 
> And parsing required ex and vi commands better way. But I was hoping
> you can use something from bash implementation perhaps.

I haven't opened the command history editing can of worms yet (come close a
couple times but bug report du jour took precedence), but I've been poking at
the getline() buffer collating recently.

(Today's blog entry at https://landley.net/notes-2023.html#03-02-2023 was like
1/3 about that. And another 1/3 ranting at glibc being tanted with mandatory
obeisance to Stallman or you don't get symbols, and bionic facepalmingly _also_
wanting #ifdef STALLMAN to give me the pipe2 syscall wrapper's prototype even
though musl and freebsd don't. But I need a portability.h thingy doing
#ifdef/#define O_DIRECT 0 anyway because macos hasn't got it, and I should be
able to pipe() and then fcntl(F_GETFL/F_SETFL) so I've got a workaround for the
prototype thing...)

Anyway, finally figured out what to implement but it's after midnight and I'm
(surprisingly!) on a day schedule at the moment. (There was an ice storm.)

The usual,

Rob