Hi
----- Original Message -----
> >
> > In some cases, it might be worth to be able to send a message with a -1
> > fd, has the protocol permits. Change add_fd/get_fd in order to track
> > if the caller wanted to send -1.
> >
> > Signed-off-by: Marc-André Lureau <marcandre.lureau@xxxxxxxxx>
> > ---
> > common/marshaller.c | 23 +++++++++++++++--------
> > common/marshaller.h | 3 ++-
> > 2 files changed, 17 insertions(+), 9 deletions(-)
> >
> > diff --git a/common/marshaller.c b/common/marshaller.c
> > index c967371..417f66b 100644
> > --- a/common/marshaller.c
> > +++ b/common/marshaller.c
> > @@ -87,6 +87,7 @@ struct SpiceMarshaller {
> > MarshallerItem *items;
> >
> > MarshallerItem static_items[N_STATIC_ITEMS];
> > + bool has_fd;
> > int fd;
>
> What about using -1 as not valid and < -1 as not filled in?
Just to get rid of a bool? that sounds more confusing to me, but really don't care that much.
>
> > };
> >
> > @@ -621,19 +622,25 @@ void *spice_marshaller_add_int8(SpiceMarshaller *m,
> > int8_t v)
> >
> > void spice_marshaller_add_fd(SpiceMarshaller *m, int fd)
> > {
> > - spice_assert(m->fd == -1);
> > + spice_assert(m->has_fd == false);
> >
> > - m->fd = dup(fd);
> > - if (m->fd == -1) {
> > - perror("dup");
> > + m->has_fd = true;
> > + if (fd != -1) {
> > + m->fd = dup(fd);
> > + if (m->fd == -1) {
> > + perror("dup");
> > + }
> > + } else {
> > + m->fd = -1;
> > }
> > }
> >
> > -int spice_marshaller_get_fd(SpiceMarshaller *m)
> > +bool spice_marshaller_get_fd(SpiceMarshaller *m, int *fd)
> > {
> > - int fd = m->fd;
> > + bool had_fd = m->has_fd;
> >
> > - m->fd = -1;
> > + *fd = m->fd;
> > + m->has_fd = false;
> >
> > - return fd;
> > + return had_fd;
> > }
> > diff --git a/common/marshaller.h b/common/marshaller.h
> > index b698b69..5b26ae4 100644
> > --- a/common/marshaller.h
> > +++ b/common/marshaller.h
> > @@ -19,6 +19,7 @@
> > #ifndef _H_MARSHALLER
> > #define _H_MARSHALLER
> >
> > +#include <stdbool.h>
> > #include <spice/macros.h>
> > #include <spice/types.h>
> > #include "mem.h"
> > @@ -67,7 +68,7 @@ void *spice_marshaller_add_int8(SpiceMarshaller *m,
> > int8_t
> > v);
> > void spice_marshaller_set_uint32(SpiceMarshaller *m, void *ref, uint32_t
> > v);
> >
> > void spice_marshaller_add_fd(SpiceMarshaller *m, int fd);
> > -int spice_marshaller_get_fd(SpiceMarshaller *m);
> > +bool spice_marshaller_get_fd(SpiceMarshaller *m, int *fd);
However, I'd keep this signature, has it allows a simpler caller to check for it
if (spice_marshaller_get_fd(m, &fd)) {
...
}
> >
> > SPICE_END_DECLS
> >
>
> Otherwise patch looks good.
> I think however is just question of style and optimization paranoia.
>
> Frediano
> _______________________________________________
> Spice-devel mailing list
> Spice-devel@xxxxxxxxxxxxxxxxxxxxx
> http://lists.freedesktop.org/mailman/listinfo/spice-devel
>
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/spice-devel