cotner@bosco.berkeley.edu (07/29/88)
A friend of mine would like to write a small C compiler for a project. (The rest of the class are writing Pascal compilers). However, he needs a formal language description of C, and so far neither one of us have been able to find one. If any one out there in netland can suggest a reference we would both appreciate it. 10Q --carl cotner@bosco.Berkeley.EDU P.S. I usually don't read this newsgroup, so please e-mail
smryan@garth.UUCP (Steven Ryan) (07/30/88)
> However, he needs >a formal language description of C, and so far neither one of us have been >able to find one. You can get a language description out of K+R, but forget about a formal definition unless you want to do it yourself.
jos@philapd.UUCP (Jos Vos) (08/01/88)
In article <12707@agate.BERKELEY.EDU> cotner@bosco.berkeley.edu writes: >A friend of mine would like to write a small C compiler for a project. >(The rest of the class are writing Pascal compilers). However, he needs >a formal language description of C, and so far neither one of us have been >able to find one. In "C: A Reference Manual", written by Harbison & Steele, a LALR(1) grammar is included in an appendix. You *should* be able to feed it to YACC, for instance, but I have never tried it! Of course (?), this is only a *syntactic* description, not a *semantic* one. -- # Jos Vos # Internet jos@philapd.UUCP # # Philips TDS, Dept SSP # # # P.O. Box 245 # UUCP ..!mcvax!philapd!jos # # 7300 AE Apeldoorn, The Netherlands # Phone +31 55 433181 #
markg@dbase.UUCP (NFS 3C501) (08/02/88)
In article <12707@agate.BERKELEY.EDU>, cotner@bosco.berkeley.edu writes: > A friend of mine would like to write a small C compiler for a project. > (The rest of the class are writing Pascal compilers). However, he needs > a formal language description of C, and so far neither one of us have been > able to find one. A more complete description of the C language is the ANSI C Standard. It is currently available only in draft form. I'm not sure where to get hold of it. You might try ANSI (American National Standards Institute) located in New York City somewhere or other on Broadway in Manhattan. They do not distribute draft versions of standards but could tell you who does distribute it.
ldg@druin.ATT.COM (Doug Gibbons) (08/03/88)
Attribute grammars are yet another type of formal language
definition, combining context free syntax with context sensitive
semantics. An MS thesis I did while at the University of Colorado
described the syntax and static semantics of C using an attribute
grammar written in Aladin. Since I was finishing just as X3J11 was
starting, the grammar does not describe ANSIisms.
Below is a grammar excerpt describing a C shift expression. The CFG
production appears first, followed by the attribution and context
conditions which must hold true. With a little effort, you can get a feel
for what is going on here. The "{in|out}" attributes are used to thread
various symbol tables through the production. The interesting things
computed here are:
SHIFT_EXP[1].at_type the expression's type
SHIFT_EXP[1].at_value the expression's value
SHIFT_EXP[1].at_eval_time when value can be known
SHIFT_EXP[1].at_const_exp is it a constant expression?
-------------------------------------------------------------------------------
rule r_39 :
SHIFT_EXP ::= SHIFT_EXP SHIFT_OP ADD_EXP
static
%
% Do the UAC balancing on the operands. Both must be integral types.
% The type is that of the converted left operand. If the right operand
% is negative or greater than the size (in bits) of the type of the
% left after the UAC, then the result is undefined.
%
SHIFT_EXP[1].at_const_exp :=
SHIFT_EXP[2].at_const_exp and ADD_EXP.at_const_exp;
SHIFT_EXP[1].at_eval_time :=
if f_is_defined(SHIFT_EXP[1].at_value) then
sc_constant
else
sc_run_time
fi;
SHIFT_EXP[1].at_id_defs_out := ADD_EXP.at_id_defs_out;
SHIFT_EXP[1].at_lval_class := sc_rval;
SHIFT_EXP[1].at_tag_names_out := ADD_EXP.at_tag_names_out;
SHIFT_EXP[1].at_tag_props_out := ADD_EXP.at_tag_props_out;
SHIFT_EXP[1].at_type :=
f_balance(SHIFT_EXP[2].at_type, ADD_EXP.at_type);
SHIFT_EXP[1].at_value :=
f_eval_binary (
SHIFT_OP.at_op,
SHIFT_EXP[2].at_value,
ADD_EXP.at_value);
SHIFT_EXP[2].at_id_defs_in := SHIFT_EXP[1].at_id_defs_in;
SHIFT_EXP[2].at_tag_names_in := SHIFT_EXP[1].at_tag_names_in;
SHIFT_EXP[2].at_tag_props_in := SHIFT_EXP[1].at_tag_props_in;
ADD_EXP.at_id_defs_in := SHIFT_EXP[2].at_id_defs_out;
ADD_EXP.at_tag_names_in := SHIFT_EXP[2].at_tag_names_out;
ADD_EXP.at_tag_props_in := SHIFT_EXP[2].at_tag_props_out;
condition
f_is_integral(SHIFT_EXP[2].at_type) and
f_is_integral(ADD_EXP.at_type)
message
"error: shift requires integral operands";
condition
not f_is_neg_i(SHIFT_EXP[1].at_value)
message
"warning: result of negative shift is non-portable";
condition
if f_is_defined(ADD_EXP.at_value) then
f_is_gt_i (
ADD_EXP.at_value,
f_mach_val (
f_mach_bit_sizeof(SHIFT_EXP[1].at_type),
sc_uint))
else
true
fi
message
"warning: magnitude of right operand of shift causes undefined value";
end;
-------------------------------------------------------------------------------
-- Doug Gibbons | ldg@druhi.ATT.COM or att!druhi!ldg
-- AT&T Bell Laboratories
-- Denver COfeldmark@hanako.stars.flab.Fujitsu.JUNET (08/04/88)
In article <458@ssp15.idca.tds.philips.nl> jos@philapd.UUCP writes: > In "C: A Reference Manual", written by Harbison & Steele, a LALR(1) grammar > is included in an appendix. You *should* be able to feed it to YACC, for > instance, but I have never tried it! I haven't looked at this book in particular, but if it is anything like what appears in K&R, it is probably possible but requires modifications. A long time ago in a place far far away we had to write a C compiler and were advised NOT to use what's in the back of K&R because YACC couldn't process it correctly. Most of us paid little attention, made some modifications which I don' t remember, and came out OK. If you have the time to search for it and understand it, look for the YACC source for the UNIX C compiler. We didn't find out it was there until we were almost finished, so it didn't help much, if you know before you start, it might be useful. :-) Mark Feldman Fujitsu Laboratories Ltd. Artificial Intelligence Laboratory feldmark@hanako.stars.flab.fujitsu.junet (Japan) feldmark%hanako.stars.flab.fujitsu.junet@uunet.uu.net (USA) -- feldmark@hanako.stars.flab.fujitsu.junet (Japan) feldmark%hanako.stars.flab.fujitsu.junet@uunet.uu.net (USA)
bts@sas.UUCP (Brian T. Schellenberger) (08/06/88)
In article <1104@garth.UUCP> smryan@garth.UUCP (Steven Ryan) writes: |> However, he needs |>a formal language description of C, and so far neither one of us have been |>able to find one. | |You can get a language description out of K+R, but forget about a formal |definition unless you want to do it yourself. Well, H&S has a good formal BNF grammar (which I have used). As for semantics, I'd recommend ANSI. It's close enough to finished to be relavent, and it's the only formal (even if it is English-like) semantic description I've ever seen of C. H&S = "Harbison & Steele" Available at local bookstores. -- --Brian, __________________________________________________ the man from |Brian T. Schellenberger ...!mcnc!rti!sas!bts Babble-On |104 Willoughby Lane work: (919) 467-8000 x7783 ____________________________|Cary, NC 27513 home: (919) 469-9389