Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

[Mike Stump]

The following reply was made to PR preprocessor/7962; it has been noted by GNATS.

From: Mike Stump <mrs@apple.com>
To: Zack Weinberg <zack@codesourcery.com>
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks
Date: Wed, 18 Sep 2002 00:41:00 -0700

 On Tuesday, September 17, 2002, at 11:17 PM, Zack Weinberg wrote:
 > In 3.3 all the case statements survive in both modes, but the
 > translation unit is rejected without -traditional-cpp.
 
 Sorry for the bug report.  I can confirm that gcc now does exactly what 
 was desired.
 The report can be closed.
 
 :-(
 
 multiline strings are evil!
 

Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

[neil]

Synopsis: -traditional's handling of unbalanced " in unused #if blocks

State-Changed-From-To: open->closed
State-Changed-By: neil
State-Changed-When: Wed Sep 18 03:19:40 2002
State-Changed-Why:
    Mistaken, as discussed.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7962

Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

[Neil Booth]

The following reply was made to PR preprocessor/7962; it has been noted by GNATS.

From: Neil Booth <neil@daikokuya.co.uk>
To: mrs@apple.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks
Date: Wed, 18 Sep 2002 08:50:11 +0100

 mrs@apple.com wrote:-
 
 > ash-2.05a$ gcc t3.c -traditional-cpp -E | grep case
 > t3.c:15:40: warning: multi-line string literals are deprecated
 >   case 0:
 >   case 2:
 
 Hmm.  tradcpp.c doesn't even have this error message so I'm confused.
 You're clearly using the ISO preprocessor somehow or other.
 How is cpp or tradcpp invoked if you specify -v?
 
 > bash-2.05a$ gcc t3.c -E | grep case
 > case 0 : 
 > case 1 : 
 > case 2 : 
 
 I find this hard to believe too: it looks like this is going through
 tradcpp rather than CPP, as CPP will warn about multi-line deprecation
 and would have sucked up case 1 in the first multi-line literal.
 -v please.
 
 Are you sure Apple hasn't removed a "!" from specs by accident?
  
 > :-(  Very confusing to people that `know' that the insides of unused #if blocks are totally discarded.
 
 I am indeed confused.  I don't have a 3.1/3.2 binary around, but I can say
 it works in HEAD (3.3).
 
 Sceptically,
 
 Neil.

Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

[Zack Weinberg]

The following reply was made to PR preprocessor/7962; it has been noted by GNATS.

From: Zack Weinberg <zack@codesourcery.com>
To: mrs@apple.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks
Date: Tue, 17 Sep 2002 23:17:50 -0700

 On Wed, Sep 18, 2002 at 05:46:52AM -0000, mrs@apple.com wrote: 
 
 > If one compiles the enclosed file with -traditional-cpp, the program
 > will return with 1, if compiled with the other cpp, it returns 0, the
 > correct value.  To understand why, examine the output of -E between
 > the two and look for case 1:
 > 
 > ash-2.05a$ gcc t3.c -traditional-cpp -E | grep case
 > t3.c:15:40: warning: multi-line string literals are deprecated
 >   case 0:
 >   case 2:
 > bash-2.05a$ gcc t3.c -E | grep case
 > case 0 : 
 > case 1 : 
 > case 2 : 
 
 Odd; my testing indicates precisely the opposite effect, with both
 2.95 and 3.2.
 
 $ gcc-2.95 -traditional-cpp -E test.c | grep case
   case 0:
   case 1:
   case 2:
 $ gcc-2.95 -E test.c | grep case
   case 0:
   case 2:
 
 $ gcc-3.2 -traditional-cpp -E test.c | grep case
   case 0:
   case 1:
   case 2:
 $ gcc-3.2 -E test.c | grep case
 test.c:15:40: warning: multi-line string literals are deprecated
 test.c:22:8: warning: extra tokens at end of #endif directive
 test.c:28:8: warning: extra tokens at end of #endif directive
   case 0:
   case 2:
 
 In 3.3 all the case statements survive in both modes, but the
 translation unit is rejected without -traditional-cpp.
 
 $ ./xgcc -B./ -traditional-cpp -E test.c | grep case
 test.c:16: warning: extra tokens at end of #endif directive
 test.c:22: warning: extra tokens at end of #endif directive
 test.c:28: warning: extra tokens at end of #endif directive
   case 0:
   case 1:
   case 2:
 
 $ ./xgcc -B./ -E test.c | grep case
 test.c:15:40: missing terminating " character
 test.c:16:8: warning: extra tokens at end of #endif directive
 test.c:21:40: missing terminating " character
 test.c:22:8: warning: extra tokens at end of #endif directive
 test.c:28:8: warning: extra tokens at end of #endif directive
   case 0:
   case 1:
   case 2:
 
 The reason for the effect as I see it is quite clear - up to 2.2,
 multiline strings exist, so everything from the first panic() to the
 second, including the first #endif, the missing case statement, and
 the next #if NOCOPY, is absorbed into the string constant.  The entire
 thing is then thrown away because it's inside a failed conditional
 block (extending from the first #if NOCOPY to the _second_ #endif).
 
 In traditional mode, however, strings silently end at the end of the
 line ("a workaround for not knowing comment boundaries when
 -traditional is used to preprocess things which are not C", to
 paraphrase the manual) so all three #endif-s are honored.  And in 3.3,
 there are no multiline strings, so the same thing happens without
 -traditional-cpp.
 
 I don't see how you got the effect you described, nor do I see what
 you want changed.  Please clarify.
 
 zw

preprocessor/7962: -traditional's handling of unbalanced " in unused #if blocks

[mrs]

>Number:         7962
>Category:       preprocessor
>Synopsis:       -traditional's handling of unbalanced " in unused #if blocks
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Sep 17 22:56:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     mrs@apple.com
>Release:        version 3.1 20020420
>Organization:
>Environment:

>Description:
If one compiles the enclosed file with -traditional-cpp, the program will return with 1, if compiled with the other cpp, it returns 0, the correct value.  To understand why, examine the output of -E between the two and look for case 1:

ash-2.05a$ gcc t3.c -traditional-cpp -E | grep case
t3.c:15:40: warning: multi-line string literals are deprecated
  case 0:
  case 2:
bash-2.05a$ gcc t3.c -E | grep case
case 0 : 
case 1 : 
case 2 : 

:-(  Very confusing to people that `know' that the insides of unused #if blocks are totally discarded.

Radar 2404967
>How-To-Repeat:

>Fix:

>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: text/plain; name="trad.c"
Content-Disposition: inline; filename="trad.c"

/* Compile with -traditional-cpp to show bug. */

#define NOCOPY 0
#include <stdio.h>

int
main()
{
  int dummy =0;

  switch (1) {

  case 0:
#if NOCOPY
    if (dummy) panic(swprint: bad dummy"); /* XXX */
#endif NOCOPY
    break;

  case 1:
#if NOCOPY
    if (dummy) panic(swprint: bad dummy"); /* XXX */
#endif NOCOPY
    return 0;

  case 2:
#if NOCOPY
     if (dummy) panic("swprint: bad dummy");
#endif NOCOPY
     break;

  default:
    return 1;
  }
}