incompatible pointer type

Started by Robert Youngover 14 years ago11 messagesbugs
Jump to latest
#1Robert Young
yayooo@gmail.com

Platform:
OpenBSD 4.9 GENERIC.MP#819 amd64 Intel(R) Xeon(R) CPU E5620 @ 2.40GHz

Error:
gmake[4]: Entering directory `/tmp/pgxc/git/src/pl/plpgsql/src'
/usr/bin/bison -d -o pl_gram.c gram.y
gcc -DPGXC -Wall -Wmissing-prototypes -Wpointer-arith
-Wdeclaration-after-statement -Wendif-labels -Wformat-security
-fno-strict-aliasing -fwrapv -fpic -DPIC -I. -I.
-I../../../../src/include -c -o pl_gram.o pl_gram.c
gram.y: In function 'plpgsql_yyparse':
gram.y:875: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:878: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1229: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1516: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1521: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1562: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1568: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1574: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1580: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1586: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1592: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1636: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1796: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1800: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:1807: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y: In function 'tok_is_keyword':
gram.y:2197: error: dereferencing pointer to incomplete type
gram.y:2197: error: dereferencing pointer to incomplete type
gram.y:2198: error: dereferencing pointer to incomplete type
gram.y: In function 'read_datatype':
gram.y:2419: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2426: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2443: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2450: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y: In function 'read_fetch_direction':
gram.y:2624: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2629: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2634: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2639: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2645: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2654: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2663: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2669: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:2674: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y: In function 'read_raise_options':
gram.y:3370: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:3373: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:3376: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gram.y:3379: warning: passing argument 2 of 'tok_is_keyword' from
incompatible pointer type
gmake[4]: *** [pl_gram.o] Error 1
gmake[4]: Leaving directory `/tmp/pgxc/git/src/pl/plpgsql/src'

CAUSE:
let's inspect an error sample:
wrong1.c :
struct bar;
void foo(struct bar *p)
{}
typedef struct
{int a;
} bar ;
bar u;
int main()
{foo(&u);
return(0);
}

# gcc wrong1.c
wrong1.c: In function 'main':
wrong1.c:9: warning: passing argument 1 of 'foo' from incompatible pointer type

BECAUSE: The former "bar" and the latter "bar" are NOT RELEVANT.
If we rename the former "bar" to "bar_anthoer" ,the compiler would
give the same reply.
wrong2.c :
struct bar_another;
void foo(struct bar_another *p)
{}
typedef struct
{int a;
} bar ;
bar u;
int main()
{foo(&u);
return(0);
}

# gcc wrong2.c
wrong2.c: In function 'main':
wrong2.c:9: warning: passing argument 1 of 'foo' from incompatible pointer type

let's fix it.
right.c :
typedef struct
{int a;
} bar ;
void foo(bar *p)
{}
bar u;
int main()
{foo(&u);
return(0);
}

FIX:
1.delete "YYSTYPE" forward-declaration
2.move "tok_is_keyword" forward-declaration behind "YYSTYPE" definition
3.delete keyword "union" in function arguments.
OK.these will eliminate all the warnings.

diff --git a/src/pl/plpgsql/src/gram.y b/src/pl/plpgsql/src/gram.y
index 4e2b705..b632a76 100644
--- a/src/pl/plpgsql/src/gram.y
+++ b/src/pl/plpgsql/src/gram.y
@@ -53,10 +53,6 @@ typedef struct

#define parser_errposition(pos) plpgsql_scanner_errposition(pos)

-union YYSTYPE;                                 /* need forward
reference for tok_is_keyword */
-
-static bool                    tok_is_keyword(int token, union YYSTYPE *lval,
-
    int kw_token, const char *kw_str);
 static void                    word_is_not_variable(PLword *word, int
location);
 static void                    cword_is_not_variable(PLcword *cword,
int location);
 static void                    current_token_is_not_variable(int tok);
@@ -318,6 +314,10 @@ static     List
*read_raise_options(void);
 %token <keyword>       K_WHEN
 %token <keyword>       K_WHILE
+%{
+static bool                    tok_is_keyword(int token, YYSTYPE *lval,
+
    int kw_token, const char *kw_str);
+%}
 %%
 pl_function            : comp_options pl_block opt_semi
@@ -2179,7 +2179,7 @@ unreserved_keyword        :
  * Hence, this kluge.
  */
 static bool
-tok_is_keyword(int token, union YYSTYPE *lval,
+tok_is_keyword(int token, YYSTYPE *lval,
                           int kw_token, const char *kw_str)
 {
        if (token == kw_token)
#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Robert Young (#1)
Re: incompatible pointer type

Robert Young <yayooo@gmail.com> writes:

Platform:
OpenBSD 4.9 GENERIC.MP#819 amd64 Intel(R) Xeon(R) CPU E5620 @ 2.40GHz

Hmm, what version of bison are you using? Because the ones I've dealt
with emit

typedef union YYSTYPE {
...
} YYSTYPE;

which makes the code correct as-is. Your proposed patch seems to me
to be making more assumptions about what bison will emit (specifically,
about the ordering of various code blocks) than what we're doing now.

regards, tom lane

#3Robert Young
yayooo@gmail.com
In reply to: Tom Lane (#2)
Re: incompatible pointer type

# /usr/bin/bison -V
bison (GNU Bison) 2.3
Written by Robert Corbett and Richard Stallman.

Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
#
From:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/bison-2.3.tgz

Sure! my pl_gram.c generated by this version of bison is :

403 typedef union
404 #line 115 "gram.y"
405 {
406 core_YYSTYPE core_yystype;
.........
454 PLpgSQL_case_when *casewhen;
455 }
456 /* Line 193 of yacc.c. */
457 #line 458 "pl_gram.c"
458 YYSTYPE;

and cause the generated pl_gram.c problematic.

Show quoted text

On Tue, Oct 18, 2011 at 18:22, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Robert Young <yayooo@gmail.com> writes:

Platform:
OpenBSD  4.9 GENERIC.MP#819 amd64 Intel(R) Xeon(R) CPU E5620 @ 2.40GHz

Hmm, what version of bison are you using?  Because the ones I've dealt
with emit

typedef union YYSTYPE {
       ...
} YYSTYPE;

which makes the code correct as-is.  Your proposed patch seems to me
to be making more assumptions about what bison will emit (specifically,
about the ordering of various code blocks) than what we're doing now.

                       regards, tom lane

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Robert Young (#3)
Re: incompatible pointer type

Robert Young <yayooo@gmail.com> writes:

On Tue, Oct 18, 2011 at 18:22, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Hmm, what version of bison are you using?

# /usr/bin/bison -V
bison (GNU Bison) 2.3
Written by Robert Corbett and Richard Stallman.

Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
#
From:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/bison-2.3.tgz

Interesting. I've got bison 2.3 on my Mac laptop, and it does what I
expect (ie, it emits "typedef union YYSTYPE ..."). So do the oldest and
newest bison versions I have handy (1.875 and 2.4.3), and both of their
manuals specify that this is the expected behavior -- see
http://www.gnu.org/s/bison/manual/html_node/Union-Decl.html
about halfway down the page.

A little bit of googling suggests that this is a bug or incompatibility
with openbsd's m4 (a tool that bison relies on):
http://comments.gmane.org/gmane.comp.parsers.bison.bugs/2708
That thread petered out without any clear resolution, but maybe you
should check for m4 updates, or try installing GNU m4.

regards, tom lane

#5Tom Lane
tgl@sss.pgh.pa.us
In reply to: Tom Lane (#4)
Re: incompatible pointer type

I wrote:

A little bit of googling suggests that this is a bug or incompatibility
with openbsd's m4 (a tool that bison relies on):
http://comments.gmane.org/gmane.comp.parsers.bison.bugs/2708

Scratch that: closer reading of the page says that the complainant was not
using some openbsd-specific copy of m4, but GNU m4 1.4.4, and that the
problem is not reproducible with newer versions of m4. So what it seems
to boil down to is "get a newer m4". Especially if you've got 1.4.4.

regards, tom lane

#6Robert Young
yayooo@gmail.com
In reply to: Tom Lane (#5)
Re: incompatible pointer type

Perfect!
I've update my m4 to version 1.4.13
from:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/m4-1.4.13.tgz
the problem solved perfectly!
Thank You !!!

# /usr/bin/gm4 --version
m4 (GNU M4) 1.4.13
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html&gt;.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Rene' Seindal.
#
# M4=/usr/bin/m4 /usr/bin/bison -d -o
pl_gram.with_OpenBSD_4.9_original_m4.c gram.y
# M4=/usr/bin/gm4 /usr/bin/bison -d -o pl_gram.with_GNU_m4_1.4.13.c gram.y
# diff -u pl_gram.with_OpenBSD_4.9_original_m4.c pl_gram.with_GNU_m4_1.4.13.c
--- pl_gram.with_OpenBSD_4.9_original_m4.c      Wed Oct 19 04:55:38 2011
+++ pl_gram.with_GNU_m4_1.4.13.c        Wed Oct 19 04:55:46 2011
@@ -400,7 +400,7 @@
 #endif
 #if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED
-typedef union
+typedef union YYSTYPE
 #line 115 "gram.y"
 {
                core_YYSTYPE                    core_yystype;
@@ -453,8 +453,8 @@
                PLpgSQL_stmt_fetch              *fetch;
                PLpgSQL_case_when               *casewhen;
 }
-/* Line 193 of yacc.c.  */
-#line 458 "pl_gram.with_OpenBSD_4.9_original_m4.c"
+/* Line 187 of yacc.c.  */
+#line 458 "pl_gram.with_GNU_m4_1.4.13.c"
        YYSTYPE;
 # define yystype YYSTYPE /* obsolescent; will be withdrawn */
 # define YYSTYPE_IS_DECLARED 1
@@ -479,7 +479,7 @@
 /* Line 216 of yacc.c.  */
-#line 483 "pl_gram.with_OpenBSD_4.9_original_m4.c"
+#line 483 "pl_gram.with_GNU_m4_1.4.13.c"

#ifdef short
# undef short
@@ -4077,7 +4077,7 @@

 /* Line 1267 of yacc.c.  */
-#line 4081 "pl_gram.with_OpenBSD_4.9_original_m4.c"
+#line 4081 "pl_gram.with_GNU_m4_1.4.13.c"
       default: break;
     }
   YY_SYMBOL_PRINT ("-> $$ =", yyr1[yyn], &yyval, &yyloc);
#
Show quoted text

On Wed, Oct 19, 2011 at 03:01, Tom Lane <tgl@sss.pgh.pa.us> wrote:

I wrote:

A little bit of googling suggests that this is a bug or incompatibility
with openbsd's m4 (a tool that bison relies on):
http://comments.gmane.org/gmane.comp.parsers.bison.bugs/2708

Scratch that: closer reading of the page says that the complainant was not
using some openbsd-specific copy of m4, but GNU m4 1.4.4, and that the
problem is not reproducible with newer versions of m4.  So what it seems
to boil down to is "get a newer m4".  Especially if you've got 1.4.4.

                       regards, tom lane

#7Tom Lane
tgl@sss.pgh.pa.us
In reply to: Robert Young (#6)
Re: incompatible pointer type

Robert Young <yayooo@gmail.com> writes:

I've update my m4 to version 1.4.13
from:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/m4-1.4.13.tgz
the problem solved perfectly!

Just for the archives' sake, can you confirm which m4 version you had
before?

regards, tom lane

#8Robert Young
yayooo@gmail.com
In reply to: Robert Young (#6)
Re: incompatible pointer type

So,I think it'd better to check the m4 compatiblity in ./configure

Show quoted text

On Wed, Oct 19, 2011 at 05:10, Robert Young <yayooo@gmail.com> wrote:

Perfect!
I've update my m4 to version 1.4.13
from:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/m4-1.4.13.tgz
the problem solved perfectly!
Thank You !!!

#9Robert Young
yayooo@gmail.com
In reply to: Tom Lane (#7)
Re: incompatible pointer type

It is hard to figure out,but I inspect the source code,told me:
http://ftp.openbsd.org/pub/OpenBSD/4.9/sys.tar.gz:/usr.bin/m4/PSD.doc/m4.ms:
Line35: m4.ms 6.3 (Berkeley) 6/5/93

Show quoted text

On Wed, Oct 19, 2011 at 05:12, Tom Lane <tgl@sss.pgh.pa.us> wrote:

Robert Young <yayooo@gmail.com> writes:

I've update my m4 to version 1.4.13
from:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/m4-1.4.13.tgz
the problem solved perfectly!

Just for the archives' sake, can you confirm which m4 version you had
before?

                       regards, tom lane

#10Robert Young
yayooo@gmail.com
In reply to: Robert Young (#6)
Re: incompatible pointer type

As I tested, if you explicit appoint a union tag, the OpenBSD m4 and
GNU m4 work identically.
And This explicit definition just corresponding to the forward-declaration.

As mentioned in
http://www.gnu.org/s/bison/manual/html_node/Union-Decl.html
This feature is a POSIX extension.
I don't know what impact to other platform.

So I suppose this should be the appropriate patch to solve this problem

diff --git a/src/pl/plpgsql/src/gram.y b/src/pl/plpgsql/src/gram.y
index 4e2b705..81a91e4 100644
--- a/src/pl/plpgsql/src/gram.y
+++ b/src/pl/plpgsql/src/gram.y
@@ -112,7 +112,7 @@ static      List
*read_raise_options(void);
 %name-prefix="plpgsql_yy"
 %locations

-%union {
+%union YYSTYPE {
core_YYSTYPE core_yystype;
/* these fields must match core_YYSTYPE: */
int ival;

Show quoted text

On 2011-10-19, Robert Young <yayooo@gmail.com> wrote:

Perfect!
I've update my m4 to version 1.4.13
from:
http://ftp.openbsd.org/pub/OpenBSD/4.9/packages/amd64/m4-1.4.13.tgz
the problem solved perfectly!
Thank You !!!

# /usr/bin/gm4 --version
m4 (GNU M4) 1.4.13
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later
<http://gnu.org/licenses/gpl.html&gt;.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Rene' Seindal.
#
# M4=/usr/bin/m4 /usr/bin/bison -d -o pl_gram.with_OpenBSD_4.9_original_m4.c gram.y
# M4=/usr/bin/gm4 /usr/bin/bison -d -o pl_gram.with_GNU_m4_1.4.13.c gram.y

#11Tom Lane
tgl@sss.pgh.pa.us
In reply to: Robert Young (#10)
Re: incompatible pointer type

Robert Young <yayooo@gmail.com> writes:

As I tested, if you explicit appoint a union tag, the OpenBSD m4 and
GNU m4 work identically.
And This explicit definition just corresponding to the forward-declaration.

As mentioned in
http://www.gnu.org/s/bison/manual/html_node/Union-Decl.html
This feature is a POSIX extension.
I don't know what impact to other platform.

So I suppose this should be the appropriate patch to solve this problem

I don't think we're going to worry too much about making plpgsql cope
with an m4 that's too old to support autoconf, which I gather is the
case from the bison bug report thread I mentioned before. If you're
going to be changing the source code or working from a git pull, you
need to have reasonably non-broken tools. We do provide pre-made bison
output in release tarballs, if you don't want to take the responsibility
of having a non-buggy bison available.

regards, tom lane