Re: [PATCH 1/2] widl: Only generate Proxy Stubs when functions have the call_as attribute

[Nikolay Sivov]

On 10.07.2017 7:43, Alistair Leslie-Hughes wrote:
> Fixes: https://bugs.winehq.org/show_bug.cgi?id=43302
> Signed-off-by: Alistair Leslie-Hughes
>
> tools/widl/header.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/tools/widl/header.c b/tools/widl/header.c
> index 7b97158..e48b488 100644
> --- a/tools/widl/header.c
> +++ b/tools/widl/header.c
> @@ -1172,7 +1172,7 @@ static void write_method_proto(FILE *header, const type_t *iface)
>    {
>      const var_t *func = stmt->u.var;
> - if (!is_local(func->attrs)) {
> - if (is_callas(func->attrs)) {
>   const char *callconv = get_attrp(func->type->attrs, ATTR_CALLCONV);
>   if (!callconv) callconv = "STDMETHODCALLTYPE";
>   /* proxy prototype */

This is wrong, 'call_as' is used to map different method to be used for
remote calls. Midl certainly does not limit prototype generation like
that. You mentioned 'odl' attribute on bug report, it's irrelevant too.

[PATCH 1/2] widl: Only generate Proxy Stubs when functions have the call_as attribute

[Huw Davies]

On Mon, Jul 10, 2017 at 08:17:26AM +0300, Nikolay Sivov wrote:
> On 10.07.2017 7:43, Alistair Leslie-Hughes wrote:
> > Fixes: https://bugs.winehq.org/show_bug.cgi?id=43302
> > Signed-off-by: Alistair Leslie-Hughes
> >
> > tools/widl/header.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/tools/widl/header.c b/tools/widl/header.c
> > index 7b97158..e48b488 100644
> > --- a/tools/widl/header.c
> > +++ b/tools/widl/header.c
> > @@ -1172,7 +1172,7 @@ static void write_method_proto(FILE *header, const type_t *iface)
> >    {
> >      const var_t *func = stmt->u.var;
> > - if (!is_local(func->attrs)) {
> > - if (is_callas(func->attrs)) {
> >   const char *callconv = get_attrp(func->type->attrs, ATTR_CALLCONV);
> >   if (!callconv) callconv = "STDMETHODCALLTYPE";
> >   /* proxy prototype */
>
> This is wrong, 'call_as' is used to map different method to be used for
> remote calls. Midl certainly does not limit prototype generation like
> that. You mentioned 'odl' attribute on bug report, it's irrelevant too.

Are you sure about this, the patch looks correct to me?

I now understand why [2/2 is needed.  The problem is the build breaks
after [1/2], so either the order needs changing or the patches need
combining.

Huw.

[PATCH 1/2] widl: Only generate Proxy Stubs when functions have the call_as attribute

[Nikolay Sivov]

On 10.07.2017 12:37, Huw Davies wrote:
> On Mon, Jul 10, 2017 at 08:17:26AM +0300, Nikolay Sivov wrote:
> > On 10.07.2017 7:43, Alistair Leslie-Hughes wrote:
> > > Fixes: https://bugs.winehq.org/show_bug.cgi?id=43302
> > > Signed-off-by: Alistair Leslie-Hughes
> > >
> > > tools/widl/header.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/tools/widl/header.c b/tools/widl/header.c
> > > index 7b97158..e48b488 100644
> > > --- a/tools/widl/header.c
> > > +++ b/tools/widl/header.c
> > > @@ -1172,7 +1172,7 @@ static void write_method_proto(FILE *header, const type_t *iface)
> > >    {
> > >      const var_t *func = stmt->u.var;
> > > - if (!is_local(func->attrs)) {
> > > - if (is_callas(func->attrs)) {
> > >   const char *callconv = get_attrp(func->type->attrs, ATTR_CALLCONV);
> > >   if (!callconv) callconv = "STDMETHODCALLTYPE";
> > >   /* proxy prototype */
> >
> > This is wrong, 'call_as' is used to map different method to be used for
> > remote calls. Midl certainly does not limit prototype generation like
> > that. You mentioned 'odl' attribute on bug report, it's irrelevant too.
>
> Are you sure about this, the patch looks correct to me?

I tested with midl 6.00.0361 and it always emits prototypes, midl
8.01.0622 by default does not, even though both versions are generating
interpreted stubs by default. Using local/call_as methods pair forces
prototype generation with default flags too, so I guess the patch is
correct for midl versions > 6.x.
> I now understand why [2/2 is needed.  The problem is the build breaks
> after [1/2], so either the order needs changing or the patches need
> combining.
>
> Huw.