On Fri, 15 Nov 2019, Markus Elfring wrote:

> > --- array-released.diff	2019-11-14 21:29:11.020576916 +0100
> > +++ array-reduced1.diff	2019-11-14 21:45:58.931956527 +0100
> > @@ -6,24 +6,10 @@
> >   	r->entry_count = t->entry_count;
> >   	r->delta_depth = t->delta_depth;
> >  -	memcpy(r->entries,t->entries,t->entry_count*sizeof(t->entries[0]));
> > -+	COPY_ARRAY(r->entries, t->entries, t->entry_count);
> > ++	memcpy(r->entries,t->entries,t->entry_count*sizeof(*(t->entries)));
> >   	release_tree_content(t);
> >   	return r;
> >   }
>
> It took a while to become more aware of software development challenges
> for the safe data processing with the semantic patch language also
> at such a source code place.
> https://github.com/git/git/blob/3edfcc65fdfc708c1c8f1d314885eecf9beb9b67/fast-import.c#L640
>
> I got the impression that the Coccinelle software is occasionally able
> to determine from the search specification “sizeof(T)” the corresponding
> data type for code like “*(t->entries)”.

It can determine the type of t->entries if it has access to the definition
of the type of t.  This type may be in a header file.  If you want
Coccinelle to be able to find this information you can use the option
--all-includes or --recursive-includes.  It will be more efficient with
the option --include-headers-for-types.

julia

> But it seems that there are circumstances to consider where the desired
> data type was not automatically determined.
> Thus the data processing  can become safer by explicitly expressing
> the case distinction for the handling of expressions.
>
> Adjusted transformation rule:
> @@
> type T;
> T* dst_ptr, src_ptr;
> T[] dst_arr, src_arr;
> expression n, x;
> @@
> -memcpy
> +COPY_ARRAY
>        (
> (       dst_ptr
> |       dst_arr
> )
>        ,
> (       src_ptr
> |       src_arr
> )
>        ,
> -       (n) * \( sizeof(T) \| sizeof(*(x)) \)
> +       n
>        )
>
>
> Regards,
> Markus
> _______________________________________________
> Cocci mailing list
> Cocci@systeme.lip6.fr
> https://systeme.lip6.fr/mailman/listinfo/cocci
>