[FFmpeg-devel] [PATCH 1/4] lavu: add simple array implementation

[wm4]

On Fri, 21 Mar 2014 01:26:28 +0100
Lukasz Marek <lukasz.m.luki2 at gmail.com> wrote:

> On 21.03.2014 00:03, Michael Niedermayer wrote:
> > On Thu, Mar 20, 2014 at 10:44:41PM +0100, Lukasz Marek wrote:
> >>>>   mem.c |   13 +++++++++++++
> >>>>   mem.h |   19 +++++++++++++++++--
> >>>>   2 files changed, 30 insertions(+), 2 deletions(-)
> >>>> 7fad1be9083fcaef4435fa0273f79bceca98821b  0001-lavu-mem-add-av_dynarray_add_nofree-function.patch
> >>>>  From 4cadb3328ba018b37c5dfe05a637b50a262151c6 Mon Sep 17 00:00:00 2001
> >>>> From: Lukasz Marek <lukasz.m.luki at gmail.com>
> >>>> Date: Tue, 25 Feb 2014 01:06:06 +0100
> >>>> Subject: [PATCH] lavu/mem: add av_dynarray_add_nofree function
> >>>>
> >>>> av_dynarray_add_nofree function have similar functionality
> >>>> as existing av_dynarray_add, but it doesn't deallocate memory
> >>>> on fails.
> >>>>
> >>>> TODO: minor bump and update doc/APIChanges
> >>>>
> >>>> Signed-off-by: Lukasz Marek <lukasz.m.luki at gmail.com>
> >>>> ---
> >>>>   libavutil/mem.c |   13 +++++++++++++
> >>>>   libavutil/mem.h |   19 +++++++++++++++++--
> >>>>   2 files changed, 30 insertions(+), 2 deletions(-)
> >>>>
> >>>> diff --git a/libavutil/mem.c b/libavutil/mem.c
> >>>> index 7206ddc..b9b5742 100644
> >>>> --- a/libavutil/mem.c
> >>>> +++ b/libavutil/mem.c
> >>>> @@ -278,6 +278,19 @@ void *av_memdup(const void *p, size_t size)
> >>>>       return ptr;
> >>>>   }
> >>>>
> >>>> +int av_dynarray_add_nofree(void *tab_ptr, int *nb_ptr, void *elem)
> >>>> +{
> >>>
> >>>> +    intptr_t *tab = *(intptr_t**)tab_ptr;
> >>>
> >>> undefined behavior (strict aliasing violation)
> >>
> >>
> >> Can you tell how it is wrong?
> >
> > yes sure
> >
> > quoting the spec:
> >
> > 7 An object shall have its stored value accessed only by an lvalue expression that has one of
> >    the following types:76)
> >    - a type compatible with the effective type of the object,
> >    - a qualified version of a type compatible with the effective type of the object,
> >    - a type that is the signed or unsigned type corresponding to the effective type of the
> >        object,
> >    - a type that is the signed or unsigned type corresponding to a qualified version of the
> >        effective type of the object,
> >    - an aggregate or union type that includes one of the aforementioned types among its
> >        members (including, recursively, a member of a subaggregate or contained union), or
> >    - a character type.
> >
> >
> > i dont think intptr_t and some pointer are compatible types
> > but maybe iam missing something
> 
> tab_ptr is pointer to pointer to array data
> so this can be replaced by something similar and (maybe) less misleading
> 
> intptr_t *tab = *(void**)tab_ptr;
> 
> This line is copied from Nicolas implementation/original implementation.
> 
> Still don't know what strict aliasing have anything to do here.

This is simply undefined behavior according to C. Basically, you can't
derference a pointer that was casted from a different pointer type.
Unless one of the exceptions applies (like unions or char*).

So, this is not legal:

	struct foo *foo_ptr;
	void *void_ptr = foo_ptr;
	*(void **)void_ptr = NULL;

unless Michael and me are interpreting the standard incorrectly.