Re: [Outreachy kernel] Coccinelle: understanding its output

["Fabio M. De Francesco" <fmdefrancesco@gmail.com>]

On Monday, April 19, 2021 8:58:24 PM CEST Julia Lawall wrote:
> On Mon, 19 Apr 2021, Fabio M. De Francesco wrote:
> > I think that I don't yet understand Coccinelle's output.
> > 
> > I've just run the following command:
> > spatch --sp-file scripts/coccinelle/free/kfree.cocci --dir drivers/
staging/
> > wlan-ng/prism2fw.c
> > 
> > It outputs something that looks like a patch:
> > 
> > HANDLING: drivers/staging/wlan-ng/prism2fw.c
> > 
> >      (ONCE) already tagged but only removed, so safe
> > 
> > diff =
> > --- drivers/staging/wlan-ng/prism2fw.c
> > +++ /tmp/cocci-output-17844-0c6c7c-prism2fw.c
> > 
> > First, what is /tmp/cocci-output-17844-0c6c7c-prism2fwc? It is not a file 
in
> > / tmp. If it is a file, where is it?
> > 
> > The output goes on with lines like the following ones:
> >   @@ -1008,8 +1005,6 @@ static int writeimage(struct wlandevice
> >   
> >         rstmsg = kzalloc(sizeof(*rstmsg), GFP_KERNEL);
> >         rwrmsg = kzalloc(sizeof(*rwrmsg), GFP_KERNEL);
> >         if (!rstmsg || !rwrmsg) {
> > 
> > -               kfree(rstmsg);
> > -               kfree(rwrmsg);
> > 
> >                 netdev_err(wlandev->netdev,
> >                 
> >                            "%s: no memory for firmware download, aborting
> > 
> > download\n",
> > 
> >                            __func__);
> > 
> > Please read the two lines with a trailing '-'... Does Coccinelle mean that
> > they should be removed? If yes, I'm not able to understand why, because I
> > think they must not be removed. Am I wrong?
> 
> Sorry.  I think that the *s are a left over from a version of the semantic
> patch that was trying to do more.  The semantic patches in the Linux
> kernel are meant to be used with make coccicheck, which would have avoided
> this problem.  The key is the lines
>
"make coccicheck" fails with the following message:

Python error: No module named coccilib.elems
coccicheck failed
make: *** [Makefile:1937: coccicheck] Error 255
> 
> virtual org
> virtual report
> 
> near the top of the file.  This means that the semantic patch should only
> be run with either the option -D org (producing emacs org mode) or the
> option -D report (producing text).  Both of these modes suppress the -
> that you are seeing.
>
Run of "spatch -D report --sp-file scripts/coccinelle/api/kfree_mismatch.cocci 
--dir drivers/staging/" fails too:

init_defs_builtins: /usr/bin/../lib/coccinelle/standard.h
Python error: No module named coccilib.elems

I think that it depends on some package I have not yet installed, or to some 
environment variable that must be set for the Python interpreter to find the  
coccili.elems module... I really don't know.

Do you have any idea for solving this problem?

Thanks,

Fabio
> 
> The - comes from the uses of * in the semantic patch, that is used to
> highlight lines of interest, where the highlighting is done by putting a -
> at the beginning of the line.  It's not meant as a suggestion to remove
> the line, just that you might want to look at it.  But in this case, as
> you noted, there is nothing interesting to see.  I will remove the *s.
> 
> julia