Re: My Flex/Bison process is stalling when trying to read a left bracket.

[Philip Herron]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Philip Herron wrote:
> address@hidden wrote:
>> The processes proceed well until the compiler generated
>> encounters an opening bracket. I get the following output...
>
>> This is the cln file Aug 2 2009 [sudo] password for joe: removing
>>  lex.yy.c rm: cannot remove `lex.yy.c': No such file or directory
>>  removing lex.yy.o rm: cannot remove `lex.yy.o': No such file or
>> directory removing parse.tab.h removing parse.tab.c removing
>> parse.tab.o ---> calling bison_program ---> calling flex_program
>> lex.l:75: negative range in character class ---> compiling lexer
>> gcc: lex.yy.c: No such file or directory gcc: no input files --->
>>  compiling parser parse.tab.c: In function yyparse:
>> parse.tab.c:1391: warning: implicit declaration of function
>> yylex ---> linking gcc: lex.yy.o: No such file or directory
>
>> V.1 started
>
>
>> project Compiler (c) 2009
>
>
>> Initialising symbol table
>
>
>> Calling yyparse(), there should be debug outputs
>
>> domain line is D0 11 0 10 domain line is D1 10 11 20 ln 97.
>> Constant line is X0 D0 1 ln 97. Constant line is X1 D1 12 ln 97.
>> Constant line is X2 D0 2 ln 97. Constant line is X3 D1 15 ln 97.
>> Constant line is X4 D0 6 ln 97. Constant line is X5 D1 20 In lex
>> 40, found an OR syntax error
>
>> Completed yyparse()
>
>> ---------------------------------------- After 'Calling
>> yyparse()...'  the compiler reads infile.i and simply confirms
>> the parsing of the domains and constants.  Things get mysterious
>> after 'In lex 40, found an OR'.  This simply confirms that the
>> first token of the predicates following the above domains and
>> constants has started. But then as you see above, the program
>> stalls with a 'syntax error'.
>
>> At this point the next token in the input.i stream is and opening
>>  bracket. But it doesn't seem to get read.  There is provision
>> for the brackets in parse.y as... %token '(' ')'        /* left
> parenthesis,
>> r parenthesis @ line 59*/ and in lex.l as... "("
> {printf("found a
>> (\n");return '(';} /*line 48 */... but it never gets to this in
>> lex.l There is a line of code at lex.l 75 which may or may not be
>>  causing a problem... address@hidden         { /* error,
> unexpected
>> char */.... it's an intended exclusion list.
>
>> I have included all the relevant files associated with this small
>>  project, and if you copy them into a directory, then calling
>> ./compile should cause execution of the shell scripts to build
>> and execute the eventual compiler which is generated.
>
>> The list of files attached are... lex.l  THE SOURCE FOR THE LEXER
>>  parse.y THE SOURCE FOR THE PARSER symtable.h THE SYMBOL TABLE
>> header file. infile.i THE SOURCE FILE TO FEED TO OUR GENERATED
>> COMPILER IT COMPRISES 2 DOMAIN LINES, 6 CONSTANTS LINES followed
>> by a LIST OF PREDICATES. cln  WHICH CLEANS OFF THE OLD OBJECT
>> FILES ETC READY FOR A NEW COMPILATION. compile WHICH PROCESSES
>> THE FLEX AND BISON SOURCES THEN LINKS, AND EXECUTES.
>
>> Once the files are into a separate directory, then simply
>> entering 'compile' will attempt to build and execute the
>> consequent files generated. You should get output same as down to
>> the -------------------- separator.
>
>> Any help is appreciated, thanks.  Joe Garvey, Cork, Ireland
>
>> ----------------------------------------------------------------------
>>
>
>
>> _______________________________________________
>> address@hidden
>> http://lists.gnu.org/mailman/listinfo/help-bison
>
> Hi
>
> Ah i see your from ireland so am I :-) .
>
> Anyways yeah looking at your lexer it wouldn't generate for me but
> i just changed that line 75 instead of hard-coding a token that
> will cause a syntax fail just stuck '.' in.
>
> Anyways your problem is probably with your parser, i am not quite
> sure what this language is or is meant to be.
>
> Is the structure having to be Domain (what ever you mean by that),
> constants and predicates.
>
> So anyways the problem is your exp: rule, you have exp can be of
> the form:
>
> ( expression ) or MAX ( number, number ) ....
>
> What you have your .i file is
>
> 'or(lt(mul(X0,X1),X2),ge(abs(sub(X3,X4)),X5))'
>
> So that is an
>
> OR ( exp )
>
> The way i would do this would be something like splitting it up
> alot,
>
> OPERATION: OR | LE | EQ ... ;
>
> predicate: OPERATION '( ' expr ')'
>
> //wait do these accept numbers or does it have to be an X0 or D0 in
>  that cause your going to want to make sure you have a hash table
> to lookup to make sure they are defined symbols before running on
> null pointers anyways..
>
> param: NUM ',' NUM | DOMAIN ',' DOMAIN | expr ',' expr ... ;
>
> expr: OPERATION '('  param ')' | OPERATION '(' expr ')'
>
> hmm... something like that i am not sure if that will work properly
>  but give it ago maby gives you something to think about :)
>
> Also you might want to check out automake to generate yourself a
> makefile instead of a compile script.
>
> To make an automake for your project would be
>
> Makefile.am:
>
> bin_PROGRAMS = parser parser_SOURCES = parse.y lex.l bla.c ....
>
> done :)
>
> Hope this helps
>
> --Phil http://redbrain.co.uk

Just looking at it again this would work better:

OPERATION: OR
                    | LE
                    | EQ
                    ...
                    ;

predicate:  OPERATION '(' param ')'
                ;


param: NUM ',' NUM
           | DOMAIN ',' DOMAIN
           | predicate ',' predicate
           ... all the mixes i guess
          ;

That might work a little nicer, i hope it is seen as a terminated rule
but give it ago :)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkp/rjwACgkQAhcOgIaQQ2FlBwCfajYl0pTeBfSsTTtD5RsGDOFR
Su4AnRgw0EcyIWHSlzCzfagToiyxjsmq
=TexD
-----END PGP SIGNATURE-----