[Swig-devel] [ swig-Bugs-3161614 ] CFFI: "-" in #defined strings

[SourceForge.net <no...@so...>]

Bugs item #3161614, was opened at 2011-01-19 12:11
Message generated for change (Comment added) made by michaelkappert
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101645&aid=3161614&group_id=1645

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: parsing
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: P.Marek (pmarek)
Assigned to: Marcelo Matus (marcelomatus)
Summary: CFFI: "-" in #defined strings

Initial Comment:
Generated string constants with a "-" are wrongly parsed as expressions.

svn_ra.h has
        #define SVN_RA_CAPABILITY_LOG_REVPROPS "log-revprops"
but swig makes this:
        (cl:defconstant SVN_RA_CAPABILITY_LOG_REVPROPS (cl:- "log" "revprops"))
which doesn't compile:
        ;   Argument X is not a NUMBER: "log"

SBCL.

Please see debian bug 607116.

----------------------------------------------------------------------

Comment By: Michael Kappert (michaelkappert)
Date: 2011-02-06 20:53

Message:
Sorry about this. Please try the appended patch. I ran the cffi test suite
this time. It didn't show any regressions.

Regards,
Michael

--
cffi.patch
--
--- swig/Source/Modules/cffi.cxx.old	2011-01-19 21:24:36.605421976 +0100
+++ swig/Source/Modules/cffi.cxx.new	2011-02-06 12:20:17.332725627 +0100
@@ -564,7 +564,12 @@
 
 int CFFI::constantWrapper(Node *n) {
   String *type = Getattr(n, "type");
-  String *converted_value = convert_literal(Getattr(n, "value"), type);
+  String *converted_value;
+  if (Strcmp(type, "p.char") == 0) {
+    converted_value = NewString(Getattr(n, "rawval"));
+  } else {
+    converted_value = convert_literal(Getattr(n, "value"), type);
+  }
   String *name = lispify_name(n, Getattr(n, "sym:name"), "'constant");
 
   if (Strcmp(name, "t") == 0 || Strcmp(name, "T") == 0)


----------------------------------------------------------------------

Comment By: William Fulton (wsfulton)
Date: 2011-02-04 08:53

Message:
This break other constant wrappers when run through the SWIG test-suite:
http://www.swig.org/Doc2.0/Extending.html#Extending_test_suite. For example
the array_member, the output in array_member.lisp:

(cl:defconstant BUFF_LEN 12)
now it is:
(cl:defconstant BUFF_LEN )

This patch will need some rework.

----------------------------------------------------------------------

Comment By: Michael Kappert (michaelkappert)
Date: 2011-01-19 21:55

Message:
I think the attached patch fixes the problem. Only briefly tested with the
original input and gtkenums.h.
I noticed that the CFFI modules generated keywords for enum constants.
This looks dubious?! I don't use CFFI myself, though.

--
cffi.patch
--
--- swig/Source/Modules/cffi.cxx	2011-01-19 21:24:36.605421976 +0100
+++ swig/Source/Modules/cffi.cxx.new	2011-01-19 21:35:59.217066056 +0100
@@ -564,7 +564,7 @@
 
 int CFFI::constantWrapper(Node *n) {
   String *type = Getattr(n, "type");
-  String *converted_value = convert_literal(Getattr(n, "value"), type);
+  String *converted_value = convert_literal(Getattr(n, "rawval"), type);
   String *name = lispify_name(n, Getattr(n, "sym:name"), "'constant");
 
   if (Strcmp(name, "t") == 0 || Strcmp(name, "T") == 0)
@@ -961,6 +961,11 @@
    try_to_split defaults to true (see stub above).
 */
 String *CFFI::convert_literal(String *literal, String *type, bool
try_to_split) {
+  if ( literal == NULL ) return 0;
+  if ( *Char(literal) == '\'' || *Char(literal) == '"') {
+   return  NewString(literal);
+  }
+
   String *num_param = Copy(literal);
   String *trimmed = trim(num_param);
   String *num = strip_parens(trimmed), *res = 0;
--
cffi.patch end
--

Regards
Michael

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101645&aid=3161614&group_id=1645