Re: [PATCH 10/11] widl: Set pointer_size according to host's pointer-size.

[Alexandre Julliard]

Kai Tietz  writes:
> diff --git a/tools/widl/widl.c b/tools/widl/widl.c
> index 5df99fc..6d8ec87 100644
> --- a/tools/widl/widl.c
> +++ b/tools/widl/widl.c
> @@ -139,7 +139,7 @@ int line_number = 1;
>
> static FILE *idfile;
>
> -unsigned int pointer_size = 0;
> +unsigned int pointer_size = sizeof(void*);
>  syskind_t typelib_kind = sizeof(void*) == 8 ? SYS_WIN64 : SYS_WIN32;

This is wrong, the host pointer size shouldn't have any influence on the
generated code.

[PATCH 10/11] widl: Set pointer_size according to host's pointer-size.

[Kai Tietz]

2013/8/19 Alexandre Julliard :
> Kai Tietz  writes:
> > diff --git a/tools/widl/widl.c b/tools/widl/widl.c
> > index 5df99fc..6d8ec87 100644
> > --- a/tools/widl/widl.c
> > +++ b/tools/widl/widl.c
> > @@ -139,7 +139,7 @@ int line_number = 1;
> >
> > static FILE *idfile;
> >
> > -unsigned int pointer_size = 0;
> > +unsigned int pointer_size = sizeof(void*);
> >  syskind_t typelib_kind = sizeof(void*) == 8 ? SYS_WIN64 : SYS_WIN32;
>
> This is wrong, the host pointer size shouldn't have any influence on the
> generated code.

True, nevertheless is the host's pointer-size used as default ... see
here typelib_kind variable ... so we should set pointer_size to a
proper-default-value too, aren't we?

Kai

[PATCH 10/11] widl: Set pointer_size according to host's pointer-size.

[Alexandre Julliard]

Kai Tietz  writes:
> 2013/8/19 Alexandre Julliard :
> > Kai Tietz  writes:
> > > diff --git a/tools/widl/widl.c b/tools/widl/widl.c
> > > index 5df99fc..6d8ec87 100644
> > > --- a/tools/widl/widl.c
> > > +++ b/tools/widl/widl.c
> > > @@ -139,7 +139,7 @@ int line_number = 1;
> > >
> > > static FILE *idfile;
> > >
> > > -unsigned int pointer_size = 0;
> > > +unsigned int pointer_size = sizeof(void*);
> > >  syskind_t typelib_kind = sizeof(void*) == 8 ? SYS_WIN64 : SYS_WIN32;
> >
> > This is wrong, the host pointer size shouldn't have any influence on the
> > generated code.
>
> True, nevertheless is the host's pointer-size used as default ... see
> here typelib_kind variable ... so we should set pointer_size to a
> proper-default-value too, aren't we?

Typelibs are either 32-bit or 64-bit, so we have to pick one. Generated
C code should be identical no matter the host platform, and pointer_size
is set to 0 to enable catching places that depend on it and shouldn't.

What are you trying to fix?

[PATCH 10/11] widl: Set pointer_size according to host's pointer-size.

[Kai Tietz]

2013/8/19 Alexandre Julliard :
> Kai Tietz  writes:
> > 2013/8/19 Alexandre Julliard :
> > > Kai Tietz  writes:
> > > > diff --git a/tools/widl/widl.c b/tools/widl/widl.c
> > > > index 5df99fc..6d8ec87 100644
> > > > --- a/tools/widl/widl.c
> > > > +++ b/tools/widl/widl.c
> > > > @@ -139,7 +139,7 @@ int line_number = 1;
> > > >
> > > > static FILE *idfile;
> > > >
> > > > -unsigned int pointer_size = 0;
> > > > +unsigned int pointer_size = sizeof(void*);
> > > >  syskind_t typelib_kind = sizeof(void*) == 8 ? SYS_WIN64 : SYS_WIN32;
> > >
> > > This is wrong, the host pointer size shouldn't have any influence on the
> > > generated code.
> >
> > True, nevertheless is the host's pointer-size used as default ... see
> > here typelib_kind variable ... so we should set pointer_size to a
> > proper-default-value too, aren't we?
>
> Typelibs are either 32-bit or 64-bit, so we have to pick one. Generated
> C code should be identical no matter the host platform, and pointer_size
> is set to 0 to enable catching places that depend on it and shouldn't.
>
> What are you trying to fix?
>
> --
> Alexandre Julliard
> [e-mail]

Heh, well, I tried to fix a case where pointer-size is used, but not
proper initialized.  This fix I have now for some time in my repro, so
I forget about the testcase I had.  Actual I had an error about it for
imports AFAI recall.  I will see if I am still be able to reproduce
that and construct a testcase.

Kai