Re: c/6547: misleading printf '$' format

Re: c/6547: misleading printf '$' format

[jsm28]

Synopsis: misleading printf '$' format

State-Changed-From-To: open->closed
State-Changed-By: jsm28
State-Changed-When: Fri Aug 23 13:11:50 2002
State-Changed-Why:
    Fixed on mainline.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=6547

Re: c/6547: misleading printf '$' format

[Joseph S. Myers]

The following reply was made to PR c/6547; it has been noted by GNATS.

From: "Joseph S. Myers" <jsm28@cam.ac.uk>
To: Pierre-Canalsat PETIT <pierrecanalsat.petit.canalsat@canal-plus.com>
Cc: <gcc-bugs@gcc.gnu.org>,  <gcc-gnats@gcc.gnu.org>, 
     <gcc-patches@gcc.gnu.org>,  <peio@blutch.dyndns.org>, 
     <prpetit@canal-plus.com>
Subject: Re: c/6547: misleading printf '$' format
Date: Fri, 3 May 2002 21:19:08 +0100 (BST)

 On Fri, 3 May 2002, Pierre-Canalsat PETIT wrote:
 
 > int main(int argc, char *argv[]) {
 >   printf("Usage: %1$s [-n <option 1>] [-m <option 2>]\n"
 >        "       %3$*2$c [-i <input>] [-o <output>] [-c <nb>] [-t <ms>]\n"
 >        "       %3$*2$c <file.in> <file.out>\n", *argv, strlen(*argv), ' ');
 >   return 0;
 > }
 > 
 > This returns me a "warning: too few arguments for format".
 > But if you remove the second line of the format, it does compile
 > without warning...
 
 Thanks for that bug report.  Patch below applied to mainline only
 (since it's not a regression).  Bootstrapped with no regressions on
 i686-pc-linux-gnu.
 
 2002-05-03  Joseph S. Myers  <jsm28@cam.ac.uk>
 
 	* c-format.c (check_format_info_main): Don't check for presence of
 	parameter for * width until after operand number has been read,
 	and only check for it if format parameters are available.
 	Fixes PR c/6547.
 
 testsuite:
 2002-05-03  Joseph S. Myers  <jsm28@cam.ac.uk>
 
 	* gcc.dg/format/xopen-2.c: New test.
 
 --- c-format.c.orig	2002-04-19 19:10:11.000000000 +0000
 +++ c-format.c	2002-05-03 15:01:13.000000000 +0000
 @@ -1751,11 +1751,6 @@ check_format_info_main (status, res, inf
  	      /* "...a field width...may be indicated by an asterisk.
  		 In this case, an int argument supplies the field width..."  */
  	      ++format_chars;
 -	      if (params == 0)
 -		{
 -		  status_warning (status, "too few arguments for format");
 -		  return;
 -		}
  	      if (has_operand_number != 0)
  		{
  		  int opnum;
 @@ -1775,6 +1770,11 @@ check_format_info_main (status, res, inf
  		}
  	      if (info->first_arg_num != 0)
  		{
 +		  if (params == 0)
 +		    {
 +		      status_warning (status, "too few arguments for format");
 +		      return;
 +		    }
  		  cur_param = TREE_VALUE (params);
  		  if (has_operand_number <= 0)
  		    {
 --- testsuite/gcc.dg/format/xopen-2.c	2001-03-26 23:57:02.000000000 +0000
 +++ testsuite/gcc.dg/format/xopen-2.c	2002-05-03 14:59:20.000000000 +0000
 @@ -0,0 +1,21 @@
 +/* Test for X/Open format extensions, as found in the
 +   Single Unix Specification.  Test for bug reported by
 +   Pierre-Canalsat PETIT <pierrecanalsat.petit.canalsat@canal-plus.com>
 +   in PR c/6547.  The test for absence of a parameter for a * width was done
 +   too early in the case of operand numbers and vprintf formats.
 +*/
 +/* Origin: Joseph Myers <jsm28@cam.ac.uk> */
 +/* { dg-do compile } */
 +/* { dg-options "-std=gnu99 -Wformat" } */
 +
 +#include "format.h"
 +
 +void vbar (va_list, const char *) __attribute__((__format__(__printf__, 2, 0)));
 +
 +void
 +foo (int i, int j, va_list va)
 +{
 +  printf("%2$*1$c", i, j);
 +  printf("%2$*1$c %2$*1$c", i, j); /* { dg-bogus "too few" "bogus too few dollar" } */
 +  vbar(va, "%*s"); /* { dg-bogus "too few" "bogus too few vprintf" } */
 +}
 
 -- 
 Joseph S. Myers
 jsm28@cam.ac.uk
 

Re: c/6547: misleading printf '$' format

[Pierre-Canalsat PETIT]

The following reply was made to PR c/6547; it has been noted by GNATS.

From: "Pierre-Canalsat PETIT" <pierrecanalsat.petit.canalsat@canal-plus.com>
To: jsm28@cam.ac.uk
Cc: gcc-bugs@gcc.gnu.org,
	gcc-gnats@gcc.gnu.org,
	Joseph Myers <jsm28@srcf.ucam.org>,
	peio@blutch.dyndns.org,
	prpetit@canal-plus.com
Subject: Re: c/6547: misleading printf '$' format
Date: Fri, 3 May 2002 16:18:13 +0200

 > You can't mix $ and non-$ formats in one format string; see the Single
 > Unix Specification.  Once a non-$ format is detected, $ operand numbers
 > are no longer checked for.  Note that "%." can only be the start of a
 > non-$ format.
 
 ====(forgot to answer to all...)====
 
 True.. I forgot that, thanks.
 So gcc-3.0 has no misleading warning, good point
 (maybe a more verbose warning should be written..).
 
 Note v2.95.4 still has one :
 
 #include <stdio.h>
 int main(int argc, char *argv[]) {
   printf("%2$.*1$s\n", argc, *argv);
   return 0;
 }
 
 This gives a "warning: unknown conversion type character '1' in format"
 
 ====(Snap)====
 
 There is still a problem it seems on gcc v3.0.4.. :
 
 #include <stdio.h>
 #include <string.h>
 
 int main(int argc, char *argv[]) {
   printf("Usage: %1$s [-n <option 1>] [-m <option 2>]\n"
        "       %3$*2$c [-i <input>] [-o <output>] [-c <nb>] [-t <ms>]\n"
        "       %3$*2$c <file.in> <file.out>\n", *argv, strlen(*argv), ' ');
   return 0;
 }
 
 This returns me a "warning: too few arguments for format".
 But if you remove the second line of the format, it does compile
 without warning...
 
 --
 Pierre PETIT
 
 

Re: c/6547: misleading printf '$' format

[Joseph S. Myers]

The following reply was made to PR c/6547; it has been noted by GNATS.

From: "Joseph S. Myers" <jsm28@cam.ac.uk>
To: <peio@blutch.dyndns.org>
Cc: <gcc-gnats@gcc.gnu.org>,  <prpetit@canal-plus.com>, 
     <gcc-bugs@gcc.gnu.org>
Subject: Re: c/6547: misleading printf '$' format
Date: Fri, 3 May 2002 14:47:23 +0100 (BST)

 On 3 May 2002 peio@blutch.dyndns.org wrote:
 
 > With both v2.95 and 3.0, there is a warning on "%.*1$s":
 > "warning: unknown conversion type character '1' in format"
 > >How-To-Repeat:
 > #include <stdio.h>
 > 
 > int main (int argc, char *argv[]) {
 >   printf("%.*s == %2$.*1$s\n", argc, *argv);
 
 You can't mix $ and non-$ formats in one format string; see the Single
 Unix Specification.  Once a non-$ format is detected, $ operand numbers
 are no longer checked for.  Note that "%." can only be the start of a
 non-$ format.
 
 -- 
 Joseph S. Myers
 jsm28@cam.ac.uk
 

c/6547: misleading printf '$' format

[peio]

>Number:         6547
>Category:       c
>Synopsis:       misleading printf '$' format
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri May 03 06:06:01 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Pierre PETIT
>Release:        gcc version 3.0.4 (and 2.95.4)
>Organization:
>Environment:
Debian GNU/Linux (testing) x86
Linux kernel 2.4.18 - libc v2.2.5
>Description:
Formats with '$' character get a warning on gcc-3.0,
when compiling with -Wall (or at least -Wformat):
"warning: unknown conversion type character '$' in format"

With both v2.95 and 3.0, there is a warning on "%.*1$s":
"warning: unknown conversion type character '1' in format"
>How-To-Repeat:
#include <stdio.h>

int main (int argc, char *argv[]) {
  printf("%.*s == %2$.*1$s\n", argc, *argv);
  return 0;
}
>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted: