Re: [Nano-devel] [PATCH] browser: fix handling of large files

[Benno Schulenberg]

On Wed, Jul 15, 2015, at 12:31, Mike Frysinger wrote:
> The current code relies on the undefined behavior of 1<<40.  This fails
> because the compiler is free to optimize away checks like (1<<40 != 0).
> It also leads to build time warnings because "1" is an integer which is
> usually 32bits.

Yeah, I didn't like those warnings either, but it didn't occur to me how
to fix that.

> The code also assumes that sizeof(long) >= sizeof(off_t)
> even though we specifically build in LFS mode, so off_t should be 64bits
> while long will still be 32bits on most 32bit systems.
> 
> Rather than unwind that though, replace the code with a proper off_t

Okay.

> and by re-using earlier variables so our shift checks always fit into
> 32bits.

> +         if (modifier != 'G' || result < (1 << 10))
> +             sprintf(foo, "%4ju %cB", (intmax_t)result, modifier);

The "modifier != 'G'" isn't needed, I think;  "if (result < (1 << 10))"
should be enough, no?

Benno

-- 
http://www.fastmail.com - Choose from over 50 domains or use your own