Re: mdocml: The st_size member of struct stat is off_t, which is signed, all

Re: mdocml: The st_size member of struct stat is off_t, which is signed, all

[Ingo Schwarze]

[ Moving this to tech@ ]

Hi Joerg,

Joerg Sonnenberger wrote on Tue, Mar 10, 2015 at 10:47:38PM +0100:
> On Tue, Mar 10, 2015 at 09:18:22AM -0500, schwarze@mdocml.bsd.lv wrote:

>> Log Message:
>> -----------
>> The st_size member of struct stat is off_t, which is signed,
>> all required by POSIX.  So don't compare it against against 
>> an unsigned constant.

> Well, the change truncates the value.

How so?  The actual width of size_t will hardly be narrower
than the width of off_t on any sane platform, or am i mistaken?

The values of off_t are positive, so the range of size_t
completely covers the range of off_t that can occur.

> Wouldn't it make more sense to use just use 0x7fffffff
> as constant instead? I think POSIX ensures that int is 32bit.

That seems completely equivalent to me, but maybe i'm missing
something?

Arguably, a comparison to 0x7fffffff is easier to understand
than a cast plus a comparison to (1 << 31), though - hmm...

Yours,
  Ingo
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: mdocml: The st_size member of struct stat is off_t, which is signed, all

[Joerg Sonnenberger]

On Wed, Mar 11, 2015 at 02:10:32AM +0100, Ingo Schwarze wrote:
> [ Moving this to tech@ ]
> 
> Hi Joerg,
> 
> Joerg Sonnenberger wrote on Tue, Mar 10, 2015 at 10:47:38PM +0100:
> > On Tue, Mar 10, 2015 at 09:18:22AM -0500, schwarze@mdocml.bsd.lv wrote:
> 
> >> Log Message:
> >> -----------
> >> The st_size member of struct stat is off_t, which is signed,
> >> all required by POSIX.  So don't compare it against against 
> >> an unsigned constant.
> 
> > Well, the change truncates the value.
> 
> How so?  The actual width of size_t will hardly be narrower
> than the width of off_t on any sane platform, or am i mistaken?

Most 32bit platforms are exactly like that. size_t will be 32bit, off_t
will be 64bit. So casting the off_t to size_t will rip out the higher
bits.

Joerg
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv

Re: mdocml: The st_size member of struct stat is off_t, which is signed, all

[Ingo Schwarze]

Hi Joerg,

Joerg Sonnenberger wrote on Wed, Mar 11, 2015 at 02:18:01AM +0100:
> On Wed, Mar 11, 2015 at 02:10:32AM +0100, Ingo Schwarze wrote:
>> Joerg Sonnenberger wrote on Tue, Mar 10, 2015 at 10:47:38PM +0100:
>>> On Tue, Mar 10, 2015 at 09:18:22AM -0500, schwarze@mdocml.bsd.lv wrote:

>>>> Log Message:
>>>> -----------
>>>> The st_size member of struct stat is off_t, which is signed,
>>>> all required by POSIX.  So don't compare it against against 
>>>> an unsigned constant.

>>> Well, the change truncates the value.

>> How so?  The actual width of size_t will hardly be narrower
>> than the width of off_t on any sane platform, or am i mistaken?

> Most 32bit platforms are exactly like that. size_t will be 32bit, off_t
> will be 64bit. So casting the off_t to size_t will rip out the higher
> bits.

Indeed, i misread header files.

So i changed the line in the way you proposed.

Thanks for the cluestick!
  Ingo
--
 To unsubscribe send an email to tech+unsubscribe@mdocml.bsd.lv