Re: [PATCH 0/5] glr.cc: support syntax_error exceptions

[Akim Demaille]

Hi Askar!

> Le 5 janv. 2019 à 02:11, Askar Safin <address@hidden> a écrit :
> 
>> Good catch, thanks!  I believe this patch is much better than my previous 
>> failed attempt.
> It seems it is buggy again. :)
> I checked out 20ad776a1ed2d95495b1706492e3ccf5441f9f0f from 
> https://github.com/akimd/bison . Then I built that attached example from 
> https://lists.gnu.org/archive/html/bison-patches/2019-01/msg00000.html . And 
> typed this:
> 
>    echo 2 + 2 % | ./a.out && echo Accepted || echo Rejected
> 
> And got this:
> 
>    Invalid character: %
>    4
>    Rejected
> 
> But with %skeleton "lalr1.cc" I got this:
> 
>    Invalid character: %
>    Rejected
> 
> So, it seems current solution is buggy, but I am not sure. Because I don't 
> know how GLR parsers supposed to work. Maybe current behavior is correct.

This time, it's something deeper in the implementation of glr.c itself, which I 
don't master, and am very afraid to break.  It does appear that it manages to 
run a few more reductions after an error, which is surprising: I would 
definitely call this a bug.

You can see that same behavior with a simple "1+2(" for instance.

You'll have to live with this, until a good fix is found.

BTW: your scanner for operators and parens is wrong.