[Cocci] question on parsing g_assert_cmpint() macro

[Cocci] question on parsing g_assert_cmpint() macro

[Eric Blake]

I'm trying to use the following rather-simple coccinelle patch to
simplify qemu:

@@
expression Obj, Key, E;
@@
- qdict_put(Obj, Key, qint_from_int(E));
+ qdict_put_int(Obj, Key, E);

I noticed by chance, however, that it only found some, and not all,
instances of the pattern in the file tests/check-qdict.c
(git.qemu-project.org/?p=qemu.git;a=blob;f=tests/check-qdict.c;h=81162ee).
 Running 'spatch --verbose-parsing ...' finally let me figure out why:

...
parse error
 = File "tests/check-qdict.c", line 473, column 55, charpos = 10131
  around = '==',
  whole content =     g_assert_cmpint(qdict_array_entries(dict, "foo."),
==, 0);
badcount: 44
bad: }
bad:
bad: static void qdict_array_entries_test(void)
bad: {
bad:     QDict *dict = qdict_new();
bad:
BAD:!!!!!     g_assert_cmpint(qdict_array_entries(dict, "foo."), ==, 0);
bad:
bad:     qdict_put(dict, "bar", qint_from_int(0));
bad:     qdict_put(dict, "baz.0", qint_from_int(0));
...

So the fact that the g_assert_cmpint() macro (from glib,
https://developer.gnome.org/glib/stable/glib-Testing.html#g-assert-cmpint)
has unusual semantics that make it called differently than a normal C
function (the second argument '==' is what the parser chokes on) means
that coccinelle is then ignoring the ENTIRE function where the parse
problem occurred, and missing the instance of my pattern just afterwards.

What is the trick for teaching coccinelle about what g_assert_cmpint()
expands to, and/or completely ignoring the use of that macro, so that I
don't have to manually look for spots that the cleanup missed?

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 604 bytes
Desc: OpenPGP digital signature
URL: <https://systeme.lip6.fr/pipermail/cocci/attachments/20170404/45f19db4/attachment.asc>

[Cocci] question on parsing g_assert_cmpint() macro

[Julia Lawall]

On Tue, 4 Apr 2017, Eric Blake wrote:

> I'm trying to use the following rather-simple coccinelle patch to
> simplify qemu:
>
> @@
> expression Obj, Key, E;
> @@
> - qdict_put(Obj, Key, qint_from_int(E));
> + qdict_put_int(Obj, Key, E);
>
> I noticed by chance, however, that it only found some, and not all,
> instances of the pattern in the file tests/check-qdict.c
> (git.qemu-project.org/?p=qemu.git;a=blob;f=tests/check-qdict.c;h=81162ee).
>  Running 'spatch --verbose-parsing ...' finally let me figure out why:
>
> ...
> parse error
>  = File "tests/check-qdict.c", line 473, column 55, charpos = 10131
>   around = '==',
>   whole content =     g_assert_cmpint(qdict_array_entries(dict, "foo."),
> ==, 0);
> badcount: 44
> bad: }
> bad:
> bad: static void qdict_array_entries_test(void)
> bad: {
> bad:     QDict *dict = qdict_new();
> bad:
> BAD:!!!!!     g_assert_cmpint(qdict_array_entries(dict, "foo."), ==, 0);
> bad:
> bad:     qdict_put(dict, "bar", qint_from_int(0));
> bad:     qdict_put(dict, "baz.0", qint_from_int(0));
> ...
>
> So the fact that the g_assert_cmpint() macro (from glib,
> https://developer.gnome.org/glib/stable/glib-Testing.html#g-assert-cmpint)
> has unusual semantics that make it called differently than a normal C
> function (the second argument '==' is what the parser chokes on) means
> that coccinelle is then ignoring the ENTIRE function where the parse
> problem occurred, and missing the instance of my pattern just afterwards.
>
> What is the trick for teaching coccinelle about what g_assert_cmpint()
> expands to, and/or completely ignoring the use of that macro, so that I
> don't have to manually look for spots that the cleanup missed?

Try adding

#define g_assert_cmpint(x,y,z)

to the macro definition file standard.h.  Or you can make your own macro
definition file for qemu and give it as an argument --macro-file file.h

Write back if that does not solve the problem.

julia


>
> --
> Eric Blake   eblake redhat com    +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
>

[Cocci] question on parsing g_assert_cmpint() macro

[Eric Blake]

On 04/05/2017 12:25 AM, Julia Lawall wrote:

>> What is the trick for teaching coccinelle about what g_assert_cmpint()
>> expands to, and/or completely ignoring the use of that macro, so that I
>> don't have to manually look for spots that the cleanup missed?
> 
> Try adding
> 
> #define g_assert_cmpint(x,y,z)
> 
> to the macro definition file standard.h.  Or you can make your own macro
> definition file for qemu and give it as an argument --macro-file file.h
> 
> Write back if that does not solve the problem.

Thanks. I didn't realize qemu already had a scripts/cocci-macro-file.h
designed just for this purpose, and it includes

#define g_assert_cmpint(a, op, b)   g_assert(a op b)

which indeed solves the problem.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 604 bytes
Desc: OpenPGP digital signature
URL: <https://systeme.lip6.fr/pipermail/cocci/attachments/20170405/ec5b3db3/attachment.asc>

[Cocci] question on parsing g_assert_cmpint() macro

[Julia Lawall]

On Wed, 5 Apr 2017, Eric Blake wrote:

> On 04/05/2017 12:25 AM, Julia Lawall wrote:
>
> >> What is the trick for teaching coccinelle about what g_assert_cmpint()
> >> expands to, and/or completely ignoring the use of that macro, so that I
> >> don't have to manually look for spots that the cleanup missed?
> >
> > Try adding
> >
> > #define g_assert_cmpint(x,y,z)
> >
> > to the macro definition file standard.h.  Or you can make your own macro
> > definition file for qemu and give it as an argument --macro-file file.h
> >
> > Write back if that does not solve the problem.
>
> Thanks. I didn't realize qemu already had a scripts/cocci-macro-file.h
> designed just for this purpose, and it includes
>
> #define g_assert_cmpint(a, op, b)   g_assert(a op b)

This definition looks fine too.

julia

>
> which indeed solves the problem.
>
> --
> Eric Blake   eblake redhat com    +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
>
>