Date:	Tue, 15 Apr 1997 10:28:25 +0000 ( )
From:	"Gurunandan R. Bhat" <grbhat AT unigoa DOT ernet DOT in>
To:	djgpp AT delorie DOT com
Subject:	advice on bison required
Message-Id:	<Pine.LNX.3.91.970415101304.4100A-100000@aditya.unigoa.ernet.in>
Mime-Version:	1.0

greetings,

	i need some help with bison. as in any lexer-parser 
combination, my lexer (lexyy.c generated by flex) returns to 
the parser yyparse(), a token and sets the  semantic value in
the variable yylval, which is customarily of type YYSTYPE. 
in a particular grammar, if yylval is of type double, one 
usually sets:

#define YYSTYPE double 

at the beginning of the bison source. on doing this and 
compiling with -d, bison generates a header file which is 
expected to contain all type information required by the 
lexer to successfully pass tokens and values to the parser.
the header file so generated is then #include'd in the lexer
source.

in my case, YYSTYPE = double. yet the header file generated 
by bison -d reads:
---------------------------------
#ifndef YYSTYPE
#define YYSTYPE int
#endif
#define	NUM	258
extern YYSTYPE yylval;
---------------------------------
rather than what i would (naively?) expect:      

---------------------------------
#define YYSTYPE double
#define	NUM	258
extern YYSTYPE yylval;
---------------------------------

is there a command line switch that will automagically generate 
the second header file rather than the first? while the first is 
correct in the event that i combine the bison output and the flex 
output into a single source module (YYSTYPE is correctly #define'd 
as double at the beginning of bison output xxxx_t.c), it seems 
incorrect when the grammar is compiled from separate flex and 
bison outputs (xxxx_t.c and lexyy.c) in separate source modules. 

i have so far been manually  making the change in the header file 
generated by bison before #include'ing it in the lex source, but 
i am sure it can be done automagically. the question is: how?

many thanks in advance
gurunandan
 
 

- Raw text -