[PATCH] symbolic names - a fixed version

[Alex Rozenman]

Hello Joel, Akim.

I am sending a fixed version after an additional testing I did this weekend.
I also attached a number of tiny test grammars with very corner case errors
(like eofs, wrong symbols, bad comments etc). As a result of this testing I
fixed a small issue in scan-gram.l. Rest of the patch is exactly the same.

I seriously considered to change my implementation in scan-gram.l in order
to reduce complexity and trickiness of the code, but I decided to keep it
this way.
When a trivial scanning like the following is used:
"["{id}"]" { extract_id_from_yytext; return MY_ID; }
it is unavoidable to parse the token in your "C" code, and to fix location's
columns then. I feel that FLEX itself is much more suitable for this type of
work. On the other hand, it seems impossible to write a non tricky code for
colon_id hack and its superposition with bracketed names.

Some technical questions:
Can I help you with pushing and "cherry picking"? Is there any difference
currently b/w the main branch and 2.4? Who is responsible for documentation
and how we can handle this?

-- 
Best regards,
Alex Rozenman (address@hidden).

May-30.git.patch
Description: Text Data

errs.tgz
Description: GNU Zip compressed data