[RFA] Token cleanup in c-exp.y

[RFA] Token cleanup in c-exp.y

[Keith Seitz]

[-- Attachment #1: Type: text/plain, Size: 312 bytes --]

Hi,

The attached patch "fixes" two tokens in c-exp.y, "->*" and ".*", which 
were not previously recognized as tokens.

Keith

ChangeLog
2009-08-24  Keith Seitz  <keiths@redhat.com>

	* c-exp.y (tokentab3): Add new token, ARROW_STAR. Changed all users.
	(tokentab2): Add new token, DOT_STAR. Changed all users.

[-- Attachment #2: token-fix.patch --]
[-- Type: text/plain, Size: 1535 bytes --]

Index: c-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/c-exp.y,v
retrieving revision 1.59
diff -u -p -r1.59 c-exp.y
--- c-exp.y	7 Jul 2009 21:33:48 -0000	1.59
+++ c-exp.y	24 Aug 2009 20:41:09 -0000
@@ -232,7 +232,7 @@ static int parse_number (char *, int, in
 %left '+' '-'
 %left '*' '/' '%'
 %right UNARY INCREMENT DECREMENT
-%right ARROW '.' '[' '('
+%right ARROW ARROW_STAR '.' DOT_STAR '[' '('
 %token <ssym> BLOCKNAME 
 %token <bval> FILENAME
 %type <bval> block
@@ -333,7 +333,7 @@ exp	:	exp ARROW qualified_name
 			  write_exp_elt_opcode (STRUCTOP_MPTR); }
 	;
 
-exp	:	exp ARROW '*' exp
+exp	:	exp ARROW_STAR exp
 			{ write_exp_elt_opcode (STRUCTOP_MPTR); }
 	;
 
@@ -368,7 +368,7 @@ exp	:	exp '.' qualified_name
 			  write_exp_elt_opcode (STRUCTOP_MEMBER); }
 	;
 
-exp	:	exp '.' '*' exp
+exp	:	exp DOT_STAR exp
 			{ write_exp_elt_opcode (STRUCTOP_MEMBER); }
 	;
 
@@ -1664,7 +1664,8 @@ struct token
 static const struct token tokentab3[] =
   {
     {">>=", ASSIGN_MODIFY, BINOP_RSH, 0},
-    {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0}
+    {"<<=", ASSIGN_MODIFY, BINOP_LSH, 0},
+    {"->*", ARROW_STAR, BINOP_END, 1}
   };
 
 static const struct token tokentab2[] =
@@ -1688,7 +1689,8 @@ static const struct token tokentab2[] =
     {"==", EQUAL, BINOP_END, 0},
     {"!=", NOTEQUAL, BINOP_END, 0},
     {"<=", LEQ, BINOP_END, 0},
-    {">=", GEQ, BINOP_END, 0}
+    {">=", GEQ, BINOP_END, 0},
+    {".*", DOT_STAR, BINOP_END, 0}
   };
 
 /* Identifier-like tokens.  */

Re: [RFA] Token cleanup in c-exp.y

[Tom Tromey]

>>>>> "Keith" == Keith Seitz <keiths@redhat.com> writes:

Keith> 2009-08-24  Keith Seitz  <keiths@redhat.com>
Keith> 	* c-exp.y (tokentab3): Add new token, ARROW_STAR. Changed all users.
Keith> 	(tokentab2): Add new token, DOT_STAR. Changed all users.

This is ok, thanks.

Tom

Re: [RFA] Token cleanup in c-exp.y

[Keith Seitz]

On 08/24/2009 04:05 PM, Tom Tromey wrote:
> This is ok, thanks.

Checked in. Thank you for looking.

Keith

RE: [RFA] Token cleanup in c-exp.y

[Pierre Muller]

   Just for information:
doesn't this mean that
the parser that previously accepted
  '->  *' with spaces between the arrow and the star,
will now reject such patterns?
  Is this also what a C compile would do?

The behavior of GDB did change indeed:
Old behavior:
(top-gdb) p current_objfile->*gdb_stderr
non-pointer-to-member value used in pointer-to-member construct
(top-gdb) p current_objfile-> *gdb_stderr
non-pointer-to-member value used in pointer-to-member construct

New behavior:
(top-gdb) p current_objfile->*gdb_stderr
non-pointer-to-member value used in pointer-to-member construct
(top-gdb) p current_objfile-> *gdb_stderr
A syntax error in expression, near `*gdb_stderr'.

  The problem is that I have 
no idea what it stands for...
and I never saw '->*' nor '.*' in any C sources.

  Please forgive the question if it is silly,
and remember that my knowledge of C is limited
to what I learned to be able to contribute to the GDB project...


Pierre Muller
Pascal language support maintainer for GDB

Re: [RFA] Token cleanup in c-exp.y

[Jiang Jilin]

On Tue, Aug 25, 2009 at 5:40 PM, Pierre Muller<muller@ics.u-strasbg.fr> wrote:
>   Just for information:
> doesn't this mean that
> the parser that previously accepted
>  '->  *' with spaces between the arrow and the star,
> will now reject such patterns?
>  Is this also what a C compile would do?

Personally, I think  '->*' is _not_  really a single token or
operation like '++' in ISO C Standard, so the patch is _not_ correct.


>  The problem is that I have
> no idea what it stands for...
> and I never saw '->*' nor '.*' in any C sources.

Indeed, we should never see '->*' nor '.*' in C sources, you know, we
can use '->' or '.' to operate any member of a structure directly, so
why do we need '*' to dereference it?

In compiler principle, I think '->*' should result in a syntax error.

Here is a sample:

struct a {
      int a;
      int b;
};

int main(int argc, char **argv)
{
        struct a test;
        test.a = 10;

        test.*a = 2;
}

$ gcc main.c
main.c: In function 'main':
main.c:10: error: expected identifier before '*' token

Re: [RFA] Token cleanup in c-exp.y

[Andreas Schwab]

Jiang Jilin <freephp@gmail.com> writes:

> On Tue, Aug 25, 2009 at 5:40 PM, Pierre Muller<muller@ics.u-strasbg.fr> wrote:
>>   Just for information:
>> doesn't this mean that
>> the parser that previously accepted
>>  '->  *' with spaces between the arrow and the star,
>> will now reject such patterns?
>>  Is this also what a C compile would do?
>
> Personally, I think  '->*' is _not_  really a single token or
> operation like '++' in ISO C Standard, so the patch is _not_ correct.

In C++ both `.*' and `->*' are single tokens.  In C, in a valid
expression, `*' can never follow `.' or `->'.

Andreas.

-- 
Andreas Schwab, schwab@redhat.com
GPG Key fingerprint = D4E8 DBE3 3813 BB5D FA84  5EC7 45C6 250E 6F00 984E
"And now for something completely different."

Re: [RFA] Token cleanup in c-exp.y

[Matt Rice]

On Mon, Aug 24, 2009 at 3:27 PM, Keith Seitz<keiths@redhat.com> wrote:
> Hi,
>
> The attached patch "fixes" two tokens in c-exp.y, "->*" and ".*", which were
> not previously recognized as tokens.
>

looks like it should also apply to objc-exp.y, as it too contains all
the lines removed from c-exp.y.

Re: [RFA] Token cleanup in c-exp.y

[Tom Tromey]

>>>>> "Matt" == Matt Rice <ratmice@gmail.com> writes:

>> The attached patch "fixes" two tokens in c-exp.y, "->*" and ".*", which were
>> not previously recognized as tokens.

Matt> looks like it should also apply to objc-exp.y, as it too contains all
Matt> the lines removed from c-exp.y.

objc-exp.y has diverged a lot from c-exp.y.
This isn't the only patch that hasn't gone into both.

I think it would be best to merge the two.  Ideally someone would
volunteer to be the ObjC maintainer and do this :-)  From the ChangeLogs
it looks like the current ObjC maintainer hasn't checked in a patch
since 2003.

Tom

Re: [RFA] Token cleanup in c-exp.y

[Tom Tromey]

>>>>> "Keith" == Keith Seitz <keiths@redhat.com> writes:

Keith> +    {".*", DOT_STAR, BINOP_END, 0}

After reading the follow-up comments I realized that this needs a couple
little changes.

The last field should be 1 here, and then the tokentab2 and tokentab3
logic in yylex needs the same cxx_only treatment as ident_tokens.

I suppose that the "::" operator also needs to be C++-only.

I can do this tomorrow, unless you'd prefer to do it.

Tom

Re: [RFA] Token cleanup in c-exp.y

[Keith Seitz]

[-- Attachment #1: Type: text/plain, Size: 934 bytes --]

On 08/25/2009 08:45 AM, Tom Tromey wrote:
>>>>>> "Keith" == Keith Seitz<keiths@redhat.com>  writes:
>
> Keith>  +    {".*", DOT_STAR, BINOP_END, 0}
>
> After reading the follow-up comments I realized that this needs a couple
> little changes.
>
> The last field should be 1 here, and then the tokentab2 and tokentab3
> logic in yylex needs the same cxx_only treatment as ident_tokens.

Yes, indeed. I've fixed that

> I suppose that the "::" operator also needs to be C++-only.

I don't think we can do that. "::" is also used by gdb for non-language 
constructs like filename::function and function::variable. There are 
several scattered tests on this already.

New patch attached.

Keith

2009-08-25  Keith Seitz  <keiths@redhat.com>

	* c-exp.y (tokentab3): Add new token, ARROW_STAR. Changed all users.
	(tokentab2): Add new token, DOT_STAR. Changed all users.
	(yylex): Add cxx_only check for tokentab2 and tokentab3 searches.


[-- Attachment #2: token-fix2.patch --]
[-- Type: text/plain, Size: 1659 bytes --]

Index: c-exp.y
===================================================================
RCS file: /cvs/src/src/gdb/c-exp.y,v
retrieving revision 1.60
diff -u -p -r1.60 c-exp.y
--- c-exp.y	25 Aug 2009 00:23:07 -0000	1.60
+++ c-exp.y	25 Aug 2009 17:35:52 -0000
@@ -1683,6 +1683,8 @@ static const struct token tokentab2[] =
     {"->", ARROW, BINOP_END, 0},
     {"&&", ANDAND, BINOP_END, 0},
     {"||", OROR, BINOP_END, 0},
+    /* "::" is *not* only C++: gdb overrides its meaning in several
+       different ways, e.g., 'filename'::func, function::variable.  */
     {"::", COLONCOLON, BINOP_END, 0},
     {"<<", LSH, BINOP_END, 0},
     {">>", RSH, BINOP_END, 0},
@@ -1690,7 +1692,7 @@ static const struct token tokentab2[] =
     {"!=", NOTEQUAL, BINOP_END, 0},
     {"<=", LEQ, BINOP_END, 0},
     {">=", GEQ, BINOP_END, 0},
-    {".*", DOT_STAR, BINOP_END, 0}
+    {".*", DOT_STAR, BINOP_END, 1}
   };
 
 /* Identifier-like tokens.  */
@@ -1849,6 +1851,10 @@ yylex (void)
   for (i = 0; i < sizeof tokentab3 / sizeof tokentab3[0]; i++)
     if (strncmp (tokstart, tokentab3[i].operator, 3) == 0)
       {
+	if (tokentab3[i].cxx_only
+	    && parse_language->la_language != language_cplus)
+	  break;
+
 	lexptr += 3;
 	yylval.opcode = tokentab3[i].opcode;
 	return tokentab3[i].token;
@@ -1858,6 +1864,10 @@ yylex (void)
   for (i = 0; i < sizeof tokentab2 / sizeof tokentab2[0]; i++)
     if (strncmp (tokstart, tokentab2[i].operator, 2) == 0)
       {
+	if (tokentab2[i].cxx_only
+	    && parse_language->la_language != language_cplus)
+	  break;
+
 	lexptr += 2;
 	yylval.opcode = tokentab2[i].opcode;
 	if (in_parse_field && tokentab2[i].token == ARROW)

Re: [RFA] Token cleanup in c-exp.y

[Tom Tromey]

>>>>> "Keith" == Keith Seitz <keiths@redhat.com> writes:

Tom> The last field should be 1 here, and then the tokentab2 and tokentab3
Tom> logic in yylex needs the same cxx_only treatment as ident_tokens.

Keith> Yes, indeed. I've fixed that

Thanks.

Tom> I suppose that the "::" operator also needs to be C++-only.

Keith> I don't think we can do that. "::" is also used by gdb for
Keith> non-language constructs like filename::function and
Keith> function::variable. There are several scattered tests on this already.

Thanks for looking into this and adding the comment.

Keith> New patch attached.

This is ok.

Tom

Re: [RFA] Token cleanup in c-exp.y

[Keith Seitz]

On 08/25/2009 11:07 AM, Tom Tromey wrote:
> Keith>  New patch attached.
>
> This is ok.

I've committed this.

Keith