Re: [Cocci] Automating a kmap/mem*/kunmap transformation

[Ira Weiny <ira.weiny@intel.com>]

On Sun, Nov 29, 2020 at 11:17:14AM +0100, Markus Elfring wrote:
> > 1) How do I get matches to iterate?
> 
> How do you think about possibilities for “iteration”?
> 
> 
> >	1a) What is the difference between '<+... ...+>' and '<... ...>'?
> 
> I suggest to take another look at the section “Dot variants” in the available
> software documentation.
> https://coccinelle.gitlabpages.inria.fr/website/docs/main_grammar004.html#sec7
> 
> Would you like to distinguish if you are looking for source code parts
> which can be optional?

Yes, not every file will have the pattern I'm looking for and as I explained in
my other thread the second instance of that pattern was not being matched and I
could not understand why.  In that case either '+' or no '+' should (and as I
know now does) work.  But I thought there was some secret to them I was
missing...  Sorry...

> 
> 
> > 2) Why can't I get the type declaration (type VP;) matched correctly.  It works
> >    some times but not always.  It also matches a lot of random declarations?
> 
> I imagine that it will help to point a few source code places out
> where you find the shown transformation approach improvable.
> 
> 
> > 2a) and for some reason in the more advanced script[3] it completely breaks.
> 
> How does such an undesirable failure look like?
> 
> I wonder a bit more about the detail why you would like to enclose
> the SmPL code by the construct “<... … ...>” here.

Because of the second instance of my pattern not matching I thought I needed to
specify that the pattern could appear 0 or more times.  But I now realize that
the pattern specification was wrong.

> I find that the change specification “-VP ptr;” would mean then that
> all variable declarations should be deleted for selected source files.


> 
> 
> Would you become interested in any more code alternatives for scripts
> in the semantic patch language?
> 
> Examples:
> A)
> (
> -ptr = kmap(page);
> |
> -ptr = kmap_atomic(page);
> )
> 
> 
> This SmPL disjunction for a deletion approach can be transformed
> like the following.
> 
> -ptr = \( kmap \| kmap_atomic \) (page);

Thanks yes that helps.

> 
> 
> B)
> (
> -memset(...);
> +memIra();
> |
> -memcpy(...);
> +memIra();
> )
> 
> 
> This SmPL disjunction for a replacement approach can eventually be transformed
> like the following.
> 
> -\( memset \| memcpy \) (...)
> +memIra()
>  ;

This makes sense but it is not my final goal and the final goal requires more
manipulation of the parameters to memset/memcpy.

But I am curious to learn more so I'll keep these alternatives in mind.

Thanks!
Ira

_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci