Re: [PATCH v2] tests/9pfs: Use g_autofree and g_autoptr where possible

[Christian Schoenebeck <qemu_oss@crudebyte.com>]

On Dienstag, 1. Februar 2022 11:11:37 CET Greg Kurz wrote:
> On Mon, 31 Jan 2022 18:43:22 +0100
> 
> Christian Schoenebeck <qemu_oss@crudebyte.com> wrote:
> > On Montag, 31. Januar 2022 17:39:30 CET Greg Kurz wrote:
> > > It is recommended to use g_autofree or g_autoptr as it reduces
> > > the odds of introducing memory leaks in future changes. Let's do
> > > that everywhere a glib allocation is performed.
> > > 
> > > The virtio_9p_create_local_test_dir() function needs some extra
> > > care though : the template pointer is eventually cached into the
> > > local_test_path global variable for the duration of the tests and
> > > should not be freed. Add the g_autofree annotation but negate it
> > > with g_steal_pointer() to make it clear that the pointer ownership
> > > is dropped on purpose.
> > > 
> > > Based-on:
> > > <f6602123c6f7d0d593466231b04fba087817abbd.1642879848.git.qemu_oss@crudeb
> > > yte
> > > .com> Signed-off-by: Greg Kurz <groug@kaod.org>
> > > ---
> > > v2: - fix crash with local_test_path
> > > ---
> > > 
> > >  tests/qtest/libqos/virtio-9p.c | 30 +++++++++++++++---------------
> > >  1 file changed, 15 insertions(+), 15 deletions(-)
> > > 
> > > diff --git a/tests/qtest/libqos/virtio-9p.c
> > > b/tests/qtest/libqos/virtio-9p.c index ef96ef006adc..f0ffbc583492
> > > 100644
> > > --- a/tests/qtest/libqos/virtio-9p.c
> > > +++ b/tests/qtest/libqos/virtio-9p.c
> > > @@ -37,18 +37,22 @@ static char *concat_path(const char* a, const char*
> > > b)
> > > 
> > >      return g_build_filename(a, b, NULL);
> > >  
> > >  }
> > > 
> > > -void virtio_9p_create_local_test_dir(void)
> > > +static char *make_temp_dir(char *template)
> > > 
> > >  {
> > > 
> > > -    struct stat st;
> > > -    char *pwd = g_get_current_dir();
> > > -    char *template = concat_path(pwd, "qtest-9p-local-XXXXXX");
> > > -
> > > -    local_test_path = mkdtemp(template);
> > > -    if (!local_test_path) {
> > > +    char *path = mkdtemp(template);
> > > +    if (!path) {
> > > 
> > >          g_test_message("mkdtemp('%s') failed: %s", template,
> > > 
> > > strerror(errno)); }
> > > -    g_free(pwd);
> > > +    return path;
> > > +}
> > > +
> > > +void virtio_9p_create_local_test_dir(void)
> > > +{
> > > +    g_autofree char *pwd = g_get_current_dir();
> > > +    g_autofree char *template = concat_path(pwd,
> > > "qtest-9p-local-XXXXXX");
> > > +    struct stat st;
> > > 
> > > +    local_test_path = make_temp_dir(g_steal_pointer(&template));
> > 
> > Quite noisy diff and under the bottom line, it does not fix what it
> > originally supposed to: addressing the leak of the global variable
> > 'local_test_path'. g_steal_pointer() simply makes the previous g_autofree
> > prefix pointless.
> The g_autofree and g_steal_pointer() dance is just an idiom that
> says "I'm not freeing this pointer on purpose, it will be handled
> somewhere else". No big deal. I'll split this in two patches : one
> for the leak of 'local_test_path' with a clear comment instead of
> g_autofree and one for the remaining g_auto* conversions.

Yeah, I got that, but it still adds redundant code, as if adding a constant
literal expression like "... * 1/2 * 2".

The other changes have an actual purpose: they reduce code. This one though
just adds noise without a real purpose, plus it reverts of what I just have
done with the previous patch: getting rid of the previously already existing
separate function init_local_test_path():

https://github.com/cschoenebeck/qemu/commit/072636e2a9f3a153078e6046e52ae5a330728b68

I don't care enough though to continue defending my position on this.

> > Why not just moving g_autofree from 'template' to the global variable
> > 'local_test_path' and avoid all that noise?
> 
> Because the 'cleanup' attribute cannot be set on a static variable as
> documented in GCC:
> 
> cleanup (cleanup_function)
> 
>     The cleanup attribute runs a function when the variable goes out of
>     scope. This attribute can only be applied to auto function scope
>     variables; it may not be applied to parameters or variables with
>     static storage duration. The function must take one parameter, a
>     pointer to a type compatible with the variable. The return value
>     of the function (if any) is ignored.
> 
> and trying to do so produces:
> 
> ../../tests/qtest/libqos/virtio-9p.c:32:1: error: ‘cleanup’ attribute
> ignored [-Werror=attributes] 32 | static g_autofree char *local_test_path;
> 
>       | ^~~~~~
> 
> cc1: all warnings being treated as errors
> 
> The leak of 'local_test_path' can only be addressed with a g_free()
> in its last user, i.e. virtio_9p_remove_local_test_dir().

Agreed, that would probably be the best solution for now.

Thanks!

Best regards,
Christian Schoenebeck