Hi Levente,

On Mon, 5 Nov 2018 at 13:33, Levente Uzonyi <leves at caesar.elte.hu> wrote:
>
> Hi Alistair,
>
> I have found the cause of the problem, and it's easy to work the issue
> around. Just evaluate the following in a workspace:
>
>         Scanner allowUnderscoreAsAssignment: false.
>         RBScanner initializeClassificationTable

Yep, that fixes it.  Thanks very much for tracking this down.


> The cause of the error is that RBScanner considers _ to be #special
> instead of #alphabetic when Scanner class >> #allowUnderscoreAsAssignment
> returns true.
> In Pharo this is not an issue, because underscore assignments have been
> abandoned and underscores in selectors are allowed permanently.
> But in Squeak everything can happen (so underscore can either be #special
> or #alphabetic depending on its context) and RBScanner would need some
> rewrite to support those cases.
>
> I suggest the line
>
>         Scanner allowUnderscoreAsAssignment: false
>
> be added to BuildSqueakSpurTrunkVMMakerImage.st before RB is loaded to
> avoid this issue in future images (because VMMaker code shouldn't rely
> on underscore assignments being supported).

Shall do.

Thanks again,
Alistair


> Levente
>
> On Mon, 5 Nov 2018, Alistair Grant wrote:
>
> > Hi Levente,
> >
> > On Mon, 5 Nov 2018 at 00:35, Levente Uzonyi <leves at caesar.elte.hu> wrote:
> >>
> >> On Sun, 4 Nov 2018, Alistair Grant wrote:
> >>
> >> > Hi All,
> >> >
> >> > If I attempt to execute the following:
> >> >
> >> > | rules |
> >> >
> >> > rules := RBParseTreeRewriter new.
> >> >
> >> > rules replace: 'state at: `@index put: `@expr' with: 'state at:
> >> > `@index put: `@expr asC_unsigned_int'.
> >>
> >> I presume the version you use doesn't support underscores in selectors.
> >> Does it work if you use a selector with no underscores in it?
> >
> > That's it, thanks!  (it works without the underscores)
> >
> > I hate it when I forget to include version information, sorry:
> >
> > Squeak5.3alpha
> > latest update: #18261
> > Image format 68021 (64 bit)
> >
> > AST-Core-eem.97
> > Refactoring-Changes-eem.22
> > Refactoring-Core-ul.162
> > Refactoring-Environment-eem.11
> > Refactoring-Squeak-Platform-cwp.2
> > Refactoring-Tests-Changes-lr.16
> > Refactoring-Tests-Core-lr.56
> > Refactoring-Tests-Environment-eem.5
> >
> >
> >> How did you load RB? Did you use the Metacello configuration? (it may
> >> not be up-to-date)
> >
> > All the above appear to be the latest versions on
> > http://www.squeaksource.com/rb.
> >
> > This is a vmmaker image, they were all loaded through:
> >
> > https://github.com/OpenSmalltalk/opensmalltalk-vm/blob/Cog/image/BuildSqueakSpurTrunkVMMakerImage.st
> >
> > It sounds like there may be an updated version of RB that already
> > addresses this?
> >
> > Thanks again,
> > Alistair
> >
> >
> >> Levente
> >>
> >> >
> >> > (rules executeTree: (MD5Plugin parseTreeFor:
> >> > #md5ProcessBuffer:withState:)) ifTrue:
> >> > [rules tree newSource inspect].
> >> > self error: 'parse tree rewrite failed'
> >> >
> >> >
> >> > I get "Error: End of statement list encountered" (stack trace below).
> >> > The same works and produces the expected output in Pharo 7.
> >> >
> >> > It looks like the Refactoring Browser has diverged quite a bit between
> >> > Squeak and Pharo.
> >> >
> >> > Who looks after RB in Squeak, and is there a plan to keep in sync with
> >> > Pharo, or the other way around, or should this be fixed independently?
> >> >
> >> > Thanks,
> >> > Alistair
>