[RFC, PATCH] YACC parsers

[RFC, PATCH] YACC parsers

[Mark Kettenis]

I (finally) figured out why some of the c++ tests were failing for me
on my machines.  The problem is that the c-exp.y parser ends up
calling the cp-name-parser.y parser, and the parsers generated by
yacc(4) on OpenBSD (which is essentially Berkely Yacc aka byacc) have
some global variables that conflict.  The cp-name-parser.y parser ends
up modifying the state of the c-exp.y parser, which then subsequently
crashes GDB.  This diff adds a few more defines to remap symbols to
each parser, which fixes the problem.

However, whoever origionally wrote the parsers clearly didn't think
this was the way to go:

/* Remap normal yacc parser interface names (yyparse, yylex, yyerror, etc),
   as well as gratuitiously global symbol names, so we can have multiple
   yacc generated parsers in gdb.  Note that these are only the variables
   produced by yacc.  If other parser generators (bison, byacc, etc) produce
   additional global names that conflict at link time, then those parser
   generators need to be fixed instead of adding those names to this list. */

It may not be impossible to "fix" the OpenBSD yacc(4) here, but given
the fact that it supports the -p option to add a prefix to all the
relevant symbols that would be somewhat silly.  Unfortunately,
changing the build infrastructure to use yacc -p isn't exactly
trivial.

Opinions?


2011-12-17  Mark Kettenis  <kettenis@gnu.org>
 
	* ada-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* c-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* cp-name-parser.y (yyss, yysslim, yyssp, yystacksize, yyvs) 
	(yyvsp): New defines.
	* f-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* jv-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* m2-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* objc-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.
	* p-exp.y (yyss, yysslim, yyssp, yystacksize, yyvs, yyvsp): New
	defines.

Index: ada-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/ada-exp.y,v
retrieving revision 1.44
diff -u -p -r1.44 ada-exp.y
--- ada-exp.y	11 Dec 2011 02:27:24 -0000	1.44
+++ ada-exp.y	17 Dec 2011 19:09:53 -0000
@@ -94,6 +94,12 @@
 #define yytoks	ada_toks		/* With YYDEBUG defined */
 #define yyname	ada_name		/* With YYDEBUG defined */
 #define yyrule	ada_rule		/* With YYDEBUG defined */
+#define yyss	ada_yyss
+#define yysslim	ada_yysslim
+#define yyssp	ada_yyssp
+#define yystacksize ada_yystacksize
+#define yyvs	ada_yyvs
+#define yyvsp	ada_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG	1		/* Default to yydebug support */
Index: c-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/c-exp.y,v
retrieving revision 1.84
diff -u -p -r1.84 c-exp.y
--- c-exp.y	11 Oct 2011 15:24:10 -0000	1.84
+++ c-exp.y	17 Dec 2011 19:09:54 -0000
@@ -104,6 +104,12 @@
 #define yygindex c_yygindex
 #define yytable	 c_yytable
 #define yycheck	 c_yycheck
+#define yyss	c_yyss
+#define yysslim	c_yysslim
+#define yyssp	c_yyssp
+#define yystacksize c_yystacksize
+#define yyvs	c_yyvs
+#define yyvsp	c_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG 1		/* Default to yydebug support */
Index: cp-name-parser.y
===================================================================
RCS file: /cvs/src/src/gdb/cp-name-parser.y,v
retrieving revision 1.22
diff -u -p -r1.22 cp-name-parser.y
--- cp-name-parser.y	18 Aug 2011 16:17:37 -0000	1.22
+++ cp-name-parser.y	17 Dec 2011 19:09:55 -0000
@@ -171,6 +171,12 @@ static struct demangle_component *d_bina
 #define yygindex cpname_yygindex
 #define yytable	 cpname_yytable
 #define yycheck	 cpname_yycheck
+#define yyss	cpname_yyss
+#define yysslim	cpname_yysslim
+#define yyssp	cpname_yyssp
+#define yystacksize cpname_yystacksize
+#define yyvs	cpname_yyvs
+#define yyvsp	cpname_yyvsp
 
 int yyparse (void);
 static int yylex (void);
Index: f-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/f-exp.y,v
retrieving revision 1.39
diff -u -p -r1.39 f-exp.y
--- f-exp.y	19 Apr 2011 18:04:07 -0000	1.39
+++ f-exp.y	17 Dec 2011 19:09:55 -0000
@@ -106,6 +106,12 @@
 #define yygindex f_yygindex
 #define yytable	 f_yytable
 #define yycheck	 f_yycheck
+#define yyss	f_yyss
+#define yysslim	f_yysslim
+#define yyssp	f_yyssp
+#define yystacksize f_yystacksize
+#define yyvs	f_yyvs
+#define yyvsp	f_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG	1		/* Default to yydebug support */
Index: jv-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/jv-exp.y,v
retrieving revision 1.42
diff -u -p -r1.42 jv-exp.y
--- jv-exp.y	9 Jan 2011 03:08:57 -0000	1.42
+++ jv-exp.y	17 Dec 2011 19:09:55 -0000
@@ -99,6 +99,12 @@
 #define yygindex java_yygindex
 #define yytable	 java_yytable
 #define yycheck	 java_yycheck
+#define yyss	java_yyss
+#define yysslim	java_yysslim
+#define yyssp	java_yyssp
+#define yystacksize java_yystacksize
+#define yyvs	java_yyvs
+#define yyvsp	java_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG 1		/* Default to yydebug support */
Index: m2-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/m2-exp.y,v
retrieving revision 1.35
diff -u -p -r1.35 m2-exp.y
--- m2-exp.y	18 Mar 2011 13:51:41 -0000	1.35
+++ m2-exp.y	17 Dec 2011 19:09:56 -0000
@@ -100,6 +100,12 @@
 #define yygindex m2_yygindex
 #define yytable	 m2_yytable
 #define yycheck	 m2_yycheck
+#define yyss	m2_yyss
+#define yysslim	m2_yysslim
+#define yyssp	m2_yyssp
+#define yystacksize m2_yystacksize
+#define yyvs	m2_yyvs
+#define yyvsp	m2_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG 1		/* Default to yydebug support */
Index: objc-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/objc-exp.y,v
retrieving revision 1.45
diff -u -p -r1.45 objc-exp.y
--- objc-exp.y	17 Jun 2011 20:24:22 -0000	1.45
+++ objc-exp.y	17 Dec 2011 19:09:56 -0000
@@ -102,6 +102,12 @@
 #define yygindex	objc_yygindex
 #define yytable		objc_yytable
 #define yycheck		objc_yycheck
+#define yyss		objc_yyss
+#define yysslim		objc_yysslim
+#define yyssp		objc_yyssp
+#define yystacksize	objc_yystacksize
+#define yyvs		objc_yyvs
+#define yyvsp		objc_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG	0		/* Default to no yydebug support.  */
Index: p-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/p-exp.y,v
retrieving revision 1.60
diff -u -p -r1.60 p-exp.y
--- p-exp.y	8 Sep 2011 14:54:19 -0000	1.60
+++ p-exp.y	17 Dec 2011 19:09:57 -0000
@@ -106,6 +106,12 @@
 #define yygindex pascal_yygindex
 #define yytable	 pascal_yytable
 #define yycheck	 pascal_yycheck
+#define yyss	pascal_yyss
+#define yysslim	pascal_yysslim
+#define yyssp	pascal_yyssp
+#define yystacksize pascal_yystacksize
+#define yyvs	pascal_yyvs
+#define yyvsp	pascal_yyvsp
 
 #ifndef YYDEBUG
 #define	YYDEBUG 1		/* Default to yydebug support */

Re: [RFC, PATCH] YACC parsers

[Joel Brobecker]

> It may not be impossible to "fix" the OpenBSD yacc(4) here, but given
> the fact that it supports the -p option to add a prefix to all the
> relevant symbols that would be somewhat silly.  Unfortunately,
> changing the build infrastructure to use yacc -p isn't exactly
> trivial.
> 
> Opinions?

FWIW, We are already using those #defines for other symbols, so I don't
think that adding a few is going to make things worse.

-- 
Joel

Re: [RFC, PATCH] YACC parsers

[Tom Tromey]

>>>>> "Mark" == Mark Kettenis <mark.kettenis@xs4all.nl> writes:

Mark> It may not be impossible to "fix" the OpenBSD yacc(4) here, but given
Mark> the fact that it supports the -p option to add a prefix to all the
Mark> relevant symbols that would be somewhat silly.  Unfortunately,
Mark> changing the build infrastructure to use yacc -p isn't exactly
Mark> trivial.

Mark> Opinions?

What is hard about using yacc -p?

I wouldn't mind getting rid of these hacks and requiring a -p-capable
yacc.  Or even requiring Bison; I think this would only be needed by gdb
developers anyhow.

Tom

Re: [RFC, PATCH] YACC parsers

[Mark Kettenis]

> From: Tom Tromey <tromey@redhat.com>
> Date: Mon, 19 Dec 2011 08:07:49 -0700
> 
> >>>>> "Mark" == Mark Kettenis <mark.kettenis@xs4all.nl> writes:
> 
> Mark> It may not be impossible to "fix" the OpenBSD yacc(4) here, but given
> Mark> the fact that it supports the -p option to add a prefix to all the
> Mark> relevant symbols that would be somewhat silly.  Unfortunately,
> Mark> changing the build infrastructure to use yacc -p isn't exactly
> Mark> trivial.
> 
> Mark> Opinions?
> 
> What is hard about using yacc -p?

As far as I can tell, it will require an explicit rule for each
parser, whereas currently an implicit rule is sufficient.

> I wouldn't mind getting rid of these hacks and requiring a -p-capable
> yacc.

POSIX requires the -p option, so I guess that would be fine.

> Or even requiring Bison; I think this would only be needed by gdb
> developers anyhow.

My primary development systems come with a perfectly fine yacc that
isn't bison.

Re: [RFC, PATCH] YACC parsers

[Tom Tromey]

Mark> As far as I can tell, it will require an explicit rule for each
Mark> parser, whereas currently an implicit rule is sufficient.

Yeah, I see.
That isn't such a big deal.  We can stuff the body of the rule into a
variable and reuse it in each explicit rule.

>> I wouldn't mind getting rid of these hacks and requiring a -p-capable
>> yacc.

Mark> POSIX requires the -p option, so I guess that would be fine.

Fine by me.

>> Or even requiring Bison; I think this would only be needed by gdb
>> developers anyhow.

Mark> My primary development systems come with a perfectly fine yacc that
Mark> isn't bison.

I suspect eventually I will have a reason to require bison, but we can
put that off until and if it ever happens.

Tom