Excellent - thanks for tracking this down! Nice easy fix, but potentially a hard one to find :)

I'll update the parser in Espruinotools too

Thank you!

@gfwilliams @bobrippling

I suspect this has broken something.

The clock face doesn't show - it's just blank. Clicking the hw button fastloads the launcher (dtlaunch).

Fastloading dtlaunch from clock:

SyntaxError: SyntaxError: Got REGEX expected ','
SyntaxError: SyntaxError: Got REGEX expected ','

Fastloading antonclk from dtlaunch:

SyntaxError: SyntaxError: Got REGEX expected ','
Uncaught Error: Function "time" not found!
 at line 2 col 4590 in antonclk.app.js
...var timeStr=require("locale").time(date,1);g.setFontAlign(0,0)....
                                 ^
in function "draw" called from line 3 col 383 in antonclk.app.js
...;Bangle.loadWidgets();draw();setTimeout(Bangle.drawWidgets,0...
                              ^

Long press to to reset to clock face antonclk (from antonclk):

widmessages.wid.js SyntaxError: SyntaxError: Got EOF expected '}'
widminbate.wid.js SyntaxError: SyntaxError: Got EOF expected '}'
Uncaught SyntaxError: Got ',' expected EOF
 at line 2 col 138
...econds()).substr(-2)+" "+b},dow:(a,b)=>"Sunday Monday Tuesda...
                              ^
at line 2 col 4601 in antonclk.app.js
...quire("locale").time(date,1);g.setFontAlign(0,0).setFont("Anton"...
                              ^
in function "draw" called from line 3 col 383 in antonclk.app.js
...;Bangle.loadWidgets();draw();setTimeout(Bangle.drawWidgets,0...
                              ^
> 

EDIT:

The same operations as above but without pretokenized apps:

>
>
SyntaxError: SyntaxError: Got REGEX expected ','
SyntaxError: SyntaxError: Got REGEX expected ','
SyntaxError: SyntaxError: Got REGEX expected ','
>
>
>
>
SyntaxError: SyntaxError: Got REGEX expected ','
Uncaught SyntaxError: Got UNFINISHED REGEX expected EOF
 at line 13 col 24 in antonclk.app.js
var x = g.getWidth() / 2;
                     ^
in function "draw" called from line 43 col 6 in antonclk.app.js
draw();
     ^
>
>
>
>
widmessages.wid.js SyntaxError: SyntaxError: Got EOF expected '}'
widminbate.wid.js SyntaxError: SyntaxError: Got EOF expected '}'
Uncaught SyntaxError: Got UNFINISHED REGEX expected EOF
 at line 13 col 24 in antonclk.app.js
var x = g.getWidth() / 2;
                     ^
in function "draw" called from line 43 col 6 in antonclk.app.js
draw();
     ^
>
>
>
> 

Adding on, I verified this one works: https://www.espruino.com/binaries/travis/584b5886906327d0cae5bb7e7301490731fcfa58/

But this next one breaks: https://www.espruino.com/binaries/travis/28b41fd99bf8dabb841be7a3f926a0be866aaeeb/

The commit history:

• https://github.com/espruino/Espruino/commits/master/
• https://www.espruino.com/binaries/travis/?C=M;O=D

var x = g.getWidth() / 2;

And that is why we didn't have ) listed I guess! I'll revert this and the EspruinoTools changes then

Thanks for testing and reporting this @thyttan!

Did it actually work for you on a real watch @bobrippling ? It looks like none of the desktop test cases found any issues with this, but I have now just added one

Thanks for the quick find! I didn't find an issue but I only gave it a cursory check that the interpreter could handle regex.

So to fix this I think we'll need to change how things are done - it's a large change so I'll run it by here to see what you both think first.

I think we should stop lexing a regex and instead decide whether we have one based on the parsing state, so we're not dependant on a look-behind at the previous token. My reasoning here is at a parse level it's relatively trivial to tell if we're going to have a regex or not:

x = a /
//    ^ when the parser is here, we have a bianry-/, so it's obvious it's a division

x = /
//  ^ when the parser is here, we have a unary-/, so it's either a regex or a comment.
// the lexer takes care of comments, so we know it's a regex,
// so we now step through this with the usual regex parsing code

This would mean removing LEX_REGEX (and associated code) and modifying jspeUnaryExpression to handle unary-/ and parse the regex.

What do you think?

I'm afraid this won't work, at least not without crippling performance - Espruino uses bracket counting to skip over areas of code that it doesn't want to execute. Either when parsing functions, if(false) or even just pasting code into the REPL. For example:

function hello(x) {
  return /}/.match(x);
}
hello("test");

So in your case in order to parse the function we have to be able to run over the entire parse tree each time we looked over any code, because otherwise } is parsed as a token and it thinks the function has ended.

I see two options here really:

• Add a state machine in the lexer which says if it sees if/while/for followed by ( and then a ) at the same level, then it sees a / it's got a regex. Even this is nontrivial because we'll need to ensure we clone that state when we seek the Lexer to a new position
• We modify EspruinoTools such that when it minifies and there's a regex right at the start, it brackets it - because it does feel like this kind of pattern doesn't really happen in sane, hand-written code

Dang, I had wondered what the rationale was behind the lookbehind code. I think the lexer state machine could be good, I'll take a look but it might be a while - I'm away for a bit with it being summer!