RE: [PHP-DEV] Overflow checks and integral vars comparison

Hi Rowan, > -----Original Message----- > From: Rowan Collins [mailto:rowan.collins@gmail.com] > Sent: Monday, August 24, 2015 12:29 AM > To: internals@lists.php.net > Subject: Re: [PHP-DEV] Overflow checks and integral vars comparison > > On 22/08/2015 02:38, Sherif Ramadan wrote: > > I see. So you're not actually doing overflow checks then? Because at > > the point you'd be checking this zend_long or size_t it could have > > already overflowed or wrapped. The subject may have misled me to > > understand differently. > > I think I understand the confusion: you are thinking of overflow as something > which happens *within* a type based on some operation (addition, > multiplication, etc). > > Anatol is talking about overflows which occur when casting *between* > types: a value of 2^33 can safely be passed around as a 64-bit integer, no > overflow has occurred; but attempting to cast it to a 32-bit integer will > immediately overflow the 32-bit integer. > > Since many PHP extensions are wrappers around libraries which may only deal in > 32-bit types, this cast is common, necessitating range checks like the ones > proposed. > Yep, that's the exact point. Not touching arithmetic operations as it is a complex thing. Even on 64-bit fe there are intrinsics for doing 128-bit math which could be used. ASM could be helpful ofc, however it needs much care, fe even within GCC versions. So it is big and should be another story :) Overflows in the cast can happen when passing args, comparing variables of different type or signess, or even with the numeric constants. Hence that cast (zend_long)INT_MAX for example. At the end - it's still an overflow/underflow issue. I'm going for an RFC with this to target 7.1. Thanks. anatol