[FFmpeg-devel] [BUG avutil] Armv6 crash with AV_RN32
Wed Mar 3 03:26:44 CET 2010
> Pavel Pavlov <pavel at summit-tech.ca> writes:
> What makes you say that? The ARM documentation states clearly that it
Obviously you don't use firefox, these docs do not exist for firefox users.
> It also works very well on _my_ ARM1136 CPU.
> > I'm wondering, how is that possible that OS is misconfigured (HTC
> > Touch pro 2, retail unmodified version)?
> Some idiot configured the retail version of course.
Is there a way to modify that control register? Do app have access to that control register?
Since function pointers are set in dsputil_init_armv6.c, maybe it's a good idea to check if unaligned access is available before using these function pointers?
> > Does that mean that cpu has some sort of control register to disable
> > unaligned access?
> > Doesn't make sense to me, either it's available or it's not.
> What makes sense to you is irrelevant, what the CPU actually does is.
> The CPU actually does support unaligned memory accesses.
> > http://www.keil.com/support/man/docs/armasm/armasm_cihdbfje.htm : (In
> ARMv6T2 and above unaligned access is permitted. In ARMv7 and above
> unaligned access is available (and is the default))
> > http://en.wikipedia.org/wiki/ARM_architecture : ARMv6T2 goes from
> ARM1156T2(F)-S, ARM1136J(F)-S is only ARMv6
> I think you need to choose your sources of information more carefully.
> (I would have expected better accuracy from KEIL though, them being
> owned by ARM and all.)
> > To me it means that somebody added code that will crash most of htc
> > phones, HTC been using that ARM1136 for years in most of their
> > models!
> What a shame HTC can't configure an OS properly. Sorry, nothing I can
> do about that.
Well, if there is a possibility of such misconfiguration, then I think there should be some code to detect it, otherwise people will blame any SW for crashing, not the phone manufacturer
More information about the ffmpeg-devel