Re: Set operations on bool-vectors

[Daniel Colascione <dancol@dancol.org>]

[-- Attachment #1: Type: text/plain, Size: 900 bytes --]

On 9/21/13 12:36 AM, Andreas Schwab wrote:
> Daniel Colascione <dancol@dancol.org> writes:
> 
>> +static
>> +ptrdiff_t
>> +bool_vector_payload_bytes (EMACS_INT length)
>> +{
>> +  EMACS_INT exact_needed_bytes;
>> +  EMACS_INT needed_bytes;
>> +
>> +  exact_needed_bytes = (length + CHAR_BIT - 1) / CHAR_BIT;
>> +  needed_bytes =  (exact_needed_bytes + sizeof (size_t) - 1)
>> +    / sizeof (size_t);
>> +
>> +  needed_bytes *= sizeof (size_t);
> 
> Please check for overflow. 

It can't overflow.  The initial number of bits comes from a Lisp
integer, which has a bit less range than an int.  What exactly is the
routine supposed to do if the value overflows?

> Also, the return type should be EMACS_INT,
> not ptrdiff_t.

Well, vector_nbytes works with ptrdiff_t, and the allocation code works
in EMACS_INT.  I'm not sure the distinction actually matters in this case.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]