On Mon, Mar 08, 2010 at 06:47:46PM +0100, sean finney wrote: > hey folks, > > just to throw something out there, the last time i ran into a bug like > this was a few weeks back with php, also with gcc 4.4 and also on arm. > turns out it was a case of signed integer overflow, which has undefined > consequences in runtime code.
thanks for the hint, that might make sense. > looking through the code i see comments that show some awareness of this > but also some constants which are a bit suspicious. namely, grep around > for -0x80000000 and similar, which are invalid values for signed int32's, > etc. actually, from /usr/lib/limits.h, -0x80000000 is indeed the minimum value for signed int32. a brief look at the tests failing show that this is exactly what upstream tries to test, the minimum and maximum valid signed values. whether they do this correctly or not i don't know yet. > i couldn't say that i'm confident that this actually is the problem, but > i think blaming the compiler really ought to be the last resort. i'm > somewhat half-interestedly playing around with a compile on agricola atm, > if i find anything i'll be sure to share. thanks. i've also managed to get a qemu-based armel machine, and even though it's slow as hell, at least i'm not disturbing other people. so now i'm trying to debug this, as much as i can. iustin -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
