Paul Eggert <[EMAIL PROTECTED]> wrote:
> Jim Meyering <[EMAIL PROTECTED]> writes:
>
>> - if (((size_t) -1) / 2 / s < n)
>> + if ((2 * (((size_t) -1 - 1) / 3)) / s < n)
>
> That's not quite right. As an extreme case, suppose S is
> SIZE_MAX/4 + 1 and N is 2. Then (2 * (((size_t) -1 - 1)
Bruno Haible <[EMAIL PROTECTED]> wrote:
> Jim Meyering wrote:
>> I've just changed xalloc's x2nrealloc to do n = 3n/2, rather than n *= 2,
>
> Which means that the time needed for realloc() will grow by a factor of 1.7
> on average. If it matters - haven't measured it -, I would suggest to use
>
Jim Meyering <[EMAIL PROTECTED]> writes:
> - if (((size_t) -1) / 2 / s < n)
> + if ((2 * (((size_t) -1 - 1) / 3)) / s < n)
That's not quite right. As an extreme case, suppose S is
SIZE_MAX/4 + 1 and N is 2. Then (2 * (((size_t) -1 - 1) / 3)) / S
evaluates to 2 and N will appear to be
On Friday 02 February 2007 00:52, Jim Meyering wrote:
> The length of the string being searched matters. More precisely,
> it's the offsets (into the buffer being searched) of the beginning
> and end of the match.
I see.
> I've just changed xalloc's x2nrealloc to do n = 3n/2, rather than n *= 2
Jim Meyering wrote:
> I've just changed xalloc's x2nrealloc to do n = 3n/2, rather than n *= 2,
Which means that the time needed for realloc() will grow by a factor of 1.7
on average. If it matters - haven't measured it -, I would suggest to use
- n = 2*n for n < 100,
- n = 3*n/2 for n >
For background, this thread started here:
http://thread.gmane.org/gmane.comp.gnu.coreutils.bugs/9581
Andi Kleen <[EMAIL PROTECTED]> wrote:
>> [*] Can you even get a 2GB-long string into expr?
>
> At least on Linux you can't, the kernel limits arguments and environments
> to 128K or so. There ar
