Re: [CEDET-devel] need help with python parser error messages

[David P. <dav...@wa...>]

Hi Richard,

 > I noticed the following lines in my *Messages* buffer:
 >
 >   Grammar in wisent-python.el contains 1 useless nonterminals and 1 
useless rules
 >   Grammar in wisent-python.el contains 2 shift/reduce conflicts

The first message indicates unused rules that can probably be safely
removed from the grammar.

The second message indicates that certain states of the automaton
result in two different actions (shift or reduce, reduce or reduce)
that conflict.  Normally an LALR automaton must be fully deterministic
(1 state -> 1 action).  I encourage you to read the bison's manual to
learn more on LALR parsers.  In most case shift/reduce conflicts are
not errors and can be safely ignored.  It is probably the case for the
python grammar :-)

You can do M-: (set wisent-verbose-flag t) before compiling your
grammar to get detailed informations on these issues in a
"*wisent-log*" buffer.  wisent.info should explain the content of
that buffer (see chapter "Grammar/Compiling a grammar/Conflicts",
then menu item "Grammar Debugging").

 >   *** `<no-type>' default matching spec DEDENT redefined as INDENT
 >   *** `symbol' default matching spec INDENT redefined as NAME
 >   *** `number' default matching spec NAME redefined as NUMBER_LITERAL
 >   *** `string' default matching spec NUMBER_LITERAL redefined as 
STRING_LITERAL
 >   *** `newline' default matching spec STRING_LITERAL redefined as NEWLINE

The above messages are warnings issued when building the table of
terminal tokens (from %token declarations).  You can safely ignore
them.  I will have a look at the code that builds the token table to
see if I can improve these obscure messages!

Hope it helps!

Thanks.
David