bug#766: 23.0.60; interesting warning during compilation

bug#766: 23.0.60; interesting warning during compilation

[Peter Dyballa]

Hello!

The warning from gcc 4.0.1 is:

	dired.c:1003: warning: right shift count >= width of type

       /* To allow inode numbers beyond 32 bits, separate into 2 24-bit
	 high parts and a 16-bit bottom part.  */
       EMACS_INT high_ino = s.st_ino >> 32;
       EMACS_INT low_ino  = s.st_ino & 0xffffffff;


Mac OS X 10.4.11, CVS update yesterday.

--
Greetings

   Pete

A blizzard is when it snows sideways.

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 23 Aug 2008 11:00:28 +0200
> Cc: 
> 
> The warning from gcc 4.0.1 is:
> 
> 	dired.c:1003: warning: right shift count >= width of type
> 
>        /* To allow inode numbers beyond 32 bits, separate into 2 24-bit
> 	 high parts and a 16-bit bottom part.  */
>        EMACS_INT high_ino = s.st_ino >> 32;
>        EMACS_INT low_ino  = s.st_ino & 0xffffffff;

This happens on any system whose st_ino is 32 bits or less, because
this code needs to cater to systems whose st_ino is wider than 32
bits.

If someone has suggestions how to avoid this warning without making
code ugly and unreadable, and preferably without system-dependent
#ifdef's, I'm all ears.

bug#766: 23.0.60; interesting warning during compilation

[Peter Dyballa]

Am 23.08.2008 um 16:10 schrieb Eli Zaretskii:

> This happens on any system whose st_ino is 32 bits or less, because
> this code needs to cater to systems whose st_ino is wider than 32
> bits.

Could it work to type-cast the variable first into a type wider than  
32 bits?

--
Greetings

   Pete

No project was ever completed on time and within budget.
				– Cheops Law

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> Cc: 766@emacsbugs.donarmstrong.com,
>  bug-gnu-emacs@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sat, 23 Aug 2008 16:58:20 +0200
> 
> Could it work to type-cast the variable first into a type wider than  
> 32 bits?

It could, if we could be sure that 64-bit wide integer types are
supported on all platforms.  I don't know if they are; does anyone
know?

(It looks like we use "long long" etc. vary sparingly and only on
platform-specific files, where there's no problem knowing whether it
is supported.)

bug#766: 23.0.60; interesting warning during compilation

[Peter Dyballa]

Am 23.08.2008 um 19:34 schrieb Eli Zaretskii:

>> Could it work to type-cast the variable first into a type wider than
>> 32 bits?
>
> It could, if we could be sure that 64-bit wide integer types are
> supported on all platforms.  I don't know if they are; does anyone
> know?


There aren't so many (elderly) systems left ... Besides, isn't the  
long long, a data type of at least 64 bits, defined in C, ANSI C99 to  
be exact? Actually, I think, it's sad that no compiler needs to  
follow this option!

Another option is to check whether the size of the variable is bigger  
than four bytes.

--
Greetings

   Pete

Rain is saved up in cloud banks.

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> Cc: 766@emacsbugs.donarmstrong.com,
>  bug-gnu-emacs@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sun, 24 Aug 2008 00:23:29 +0200
> 
> Another option is to check whether the size of the variable is bigger  
> than four bytes.

If you mean with sizeof, then the problem is that sizeof is not a
preprocessing-time operator, so it cannot prevent the compiler from
seeing the offending code.

bug#766: 23.0.60; interesting warning during compilation

[Peter Dyballa]

Am 24.08.2008 um 05:13 schrieb Eli Zaretskii:

> If you mean with sizeof, then the problem is that sizeof is not a
> preprocessing-time operator, so it cannot prevent the compiler from
> seeing the offending code.

You mean

	if (sizeof (s.st_ino) > 4)

introducing the shift operation will fail to keep the compiler from  
emitting the warning?


Anyway, I think Andreas Schwab's proposition to split the one wide  
shift into two shorter ones is better. And the original code relies  
on an undefined behaviour of every compiler, which might change.

--
Greetings

   Pete

No project was ever completed on time and within budget.
				– Cheops Law

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> Cc: 766@emacsbugs.donarmstrong.com,
>  bug-gnu-emacs@gnu.org
> From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
> Date: Sun, 24 Aug 2008 09:27:16 +0200
> 
> 
> Am 24.08.2008 um 05:13 schrieb Eli Zaretskii:
> 
> > If you mean with sizeof, then the problem is that sizeof is not a
> > preprocessing-time operator, so it cannot prevent the compiler from
> > seeing the offending code.
> 
> You mean
> 
> 	if (sizeof (s.st_ino) > 4)
> 
> introducing the shift operation will fail to keep the compiler from  
> emitting the warning?

Yes, it will fail.  Please try compiling this toy program on a 32-bit
machine, and you will see the same warning.

    int foo (int a)
    {
      if (sizeof (a) > 4)
	return a >> 32;
      else
	return 0;
    }

> Anyway, I think Andreas Schwab's proposition to split the one wide  
> shift into two shorter ones is better.

I agree, I just wanted to be sure that this won't bring the same
warning back at some future time.

bug#766: 23.0.60; interesting warning during compilation

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> If someone has suggestions how to avoid this warning without making
> code ugly and unreadable, and preferably without system-dependent
> #ifdef's, I'm all ears.

EMACS_INT high_ino = s.st_ino >> 31 >> 1;

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: 766@emacsbugs.donarmstrong.com,
> 	Peter Dyballa <Peter_Dyballa@Freenet.DE>, bug-gnu-emacs@gnu.org
> Date: Sat, 23 Aug 2008 19:33:22 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > If someone has suggestions how to avoid this warning without making
> > code ugly and unreadable, and preferably without system-dependent
> > #ifdef's, I'm all ears.
> 
> EMACS_INT high_ino = s.st_ino >> 31 >> 1;

Thanks.  Can we be reasonably sure that GCC will not become smart
enough some day to flag this as well?

bug#766: 23.0.60; interesting warning during compilation

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Andreas Schwab <schwab@suse.de>
>> EMACS_INT high_ino = s.st_ino >> 31 >> 1;
>
> Thanks.  Can we be reasonably sure that GCC will not become smart
> enough some day to flag this as well?

What do you mean with "smart enough"?  There is nothing to flag here.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: 766@emacsbugs.donarmstrong.com, Peter_Dyballa@Freenet.DE,
> 	bug-gnu-emacs@gnu.org
> Date: Sat, 23 Aug 2008 20:56:38 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> From: Andreas Schwab <schwab@suse.de>
> >> EMACS_INT high_ino = s.st_ino >> 31 >> 1;
> >
> > Thanks.  Can we be reasonably sure that GCC will not become smart
> > enough some day to flag this as well?
> 
> What do you mean with "smart enough"?

Maybe that was a polite way of saying "silly enough".

> There is nothing to flag here.

I meant the warning about shifting by too many bits.  31+1 = 32, and
even a compiler can realize that.

Please don't get me wrong: I will gladly make the change as you
suggested, I just don't want us to bump into this again a few GCC
upgrades down the road.

bug#766: 23.0.60; interesting warning during compilation

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> I meant the warning about shifting by too many bits.  31+1 = 32, and
> even a compiler can realize that.

That's not how the shift operator is defined.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: 766@emacsbugs.donarmstrong.com, Peter_Dyballa@Freenet.DE,
> 	bug-gnu-emacs@gnu.org
> Date: Sat, 23 Aug 2008 22:03:31 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I meant the warning about shifting by too many bits.  31+1 = 32, and
> > even a compiler can realize that.
> 
> That's not how the shift operator is defined.

Andreas, it's late and I'm tired.  Could you please answer my
question, which is: what is the danger, if any, that GCC in some
future version will produce the same warning for the code you
suggested?

Thanks in advance.

bug#766: 23.0.60; interesting warning during compilation

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> Andreas, it's late and I'm tired.  Could you please answer my
> question, which is: what is the danger, if any, that GCC in some
> future version will produce the same warning for the code you
> suggested?

I have already answered that question.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux Products GmbH, Maxfeldstraße 5, 90409 Nürnberg, Germany
PGP key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#766: 23.0.60; interesting warning during compilation

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Cc: 766@emacsbugs.donarmstrong.com, Peter_Dyballa@Freenet.DE,
> 	bug-gnu-emacs@gnu.org
> Date: Sun, 24 Aug 2008 01:12:55 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Andreas, it's late and I'm tired.  Could you please answer my
> > question, which is: what is the danger, if any, that GCC in some
> > future version will produce the same warning for the code you
> > suggested?
> 
> I have already answered that question.

In that case, I didn't understand it.  Just for me, could you please
translate it into a simple YES or NO: is there a danger that GCC will
in the future produce the same warning for your suggested code?

Thanks.

bug#766: 23.0.60; interesting warning during compilation

[Stefan Monnier]

>> > If someone has suggestions how to avoid this warning without making
>> > code ugly and unreadable, and preferably without system-dependent
>> > #ifdef's, I'm all ears.
>> 
>> EMACS_INT high_ino = s.st_ino >> 31 >> 1;

> Thanks.  Can we be reasonably sure that GCC will not become smart
> enough some day to flag this as well?

I don't think anybody can guarantee it, but usually it's a bad idea to
output warnings about post-optimization code, so you may hope that GCC
will optimize "x >> 31 >> 1" to "x >> 32" without ever outputing
a warning about the silliness of the end result.

Then again, the present code shows that even outputting a warning about
"x >> 32" is already borderline annoying, so there's no guarantee gcc
won't become more annoying in the future.


        Stefan

bug#766: marked as done (23.0.60; interesting warning during compilation)

[Emacs bug Tracking System]

[-- Attachment #1: Type: text/plain, Size: 862 bytes --]


Your message dated Wed, 27 Aug 2008 15:05:59 -0400
with message-id <87k5e2uvl4.fsf@cyd.mit.edu>
and subject line Re: bug#766: 23.0.60; interesting warning during compilation
has caused the Emacs bug report #766,
regarding 23.0.60; interesting warning during compilation
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact don@donarmstrong.com
immediately.)


-- 
766: http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=766
Emacs Bug Tracking System
Contact don@donarmstrong.com with problems

[-- Attachment #2: Type: message/rfc822, Size: 2728 bytes --]

From: Peter Dyballa <Peter_Dyballa@Freenet.DE>
To: emacs-pretest-bug@gnu.org
Subject: 23.0.60; interesting warning during compilation
Date: Sat, 23 Aug 2008 11:00:28 +0200
Message-ID: <4762603A-B23E-454E-A43C-422DE5036DC4@Freenet.DE>

Hello!

The warning from gcc 4.0.1 is:

	dired.c:1003: warning: right shift count >= width of type

       /* To allow inode numbers beyond 32 bits, separate into 2 24-bit
	 high parts and a 16-bit bottom part.  */
       EMACS_INT high_ino = s.st_ino >> 32;
       EMACS_INT low_ino  = s.st_ino & 0xffffffff;


Mac OS X 10.4.11, CVS update yesterday.

--
Greetings

   Pete

A blizzard is when it snows sideways.





[-- Attachment #3: Type: message/rfc822, Size: 967 bytes --]

From: Chong Yidong <cyd@stupidchicken.com>
To: 766-done@emacsbugs.donarmstrong.com
Subject: Re: bug#766: 23.0.60; interesting warning during compilation
Date: Wed, 27 Aug 2008 15:05:59 -0400
Message-ID: <87k5e2uvl4.fsf@cyd.mit.edu>

I've checked in Andreas' fix.