Re: oleaut32: do not crash in logging if string is NULL

[Alexandre Julliard]

Aric Stewart <aric(a)codeweavers.com> writes:
> @@ -68,10 +68,16 @@ typedef struct _marshal_state {
>
> /* used in the olerelay code to avoid having the L"" stuff added by debugstr_w */
>  static char *relaystr(WCHAR *in) {
> -    char *tmp = (char *)debugstr_w(in);
> -    tmp += 2;
> -    tmp[strlen(tmp)-1] = '\0';
> -    return tmp;
> +    static char szNull[] = "<NULL>";
> +    if (in)
> +    {
> +        char *tmp = (char *)debugstr_w(in);
> +        tmp += 2;
> +        tmp[strlen(tmp)-1] = '\0';
> +        return tmp;
> +    }
> +    else
> +        return szNull;
>  }

This should never happen. Places that can get a null pointer should use
the standard debugstr_w.

-- 
Alexandre Julliard
julliard(a)winehq.org

oleaut32: do not crash in logging if string is NULL

[Aric Stewart]

Meaning the place where I am finding this is different bug?

In deserialize_param in the VT_BSTR case if the byte_length of the BSTR 
that we are deserializing is -1  then BSTR_UserUnmarshal places NULL 
into *arg.

I will admit I know very little about this code but having it crash when 
trying to log seemed very wrong.

Should I change the
if (debugout) TRACE_(olerelay)("%s",relaystr(*(BSTR *)arg));
to a simple debugstr_w or is this revealing a deeper problem somewhere?

thanks,
-aric

Alexandre Julliard wrote:
> Aric Stewart <aric(a)codeweavers.com> writes:
> > @@ -68,10 +68,16 @@ typedef struct _marshal_state {
> >
> > /* used in the olerelay code to avoid having the L"" stuff added by debugstr_w */
> >  static char *relaystr(WCHAR *in) {
> > -    char *tmp = (char *)debugstr_w(in);
> > -    tmp += 2;
> > -    tmp[strlen(tmp)-1] = '\0';
> > -    return tmp;
> > +    static char szNull[] = "<NULL>";
> > +    if (in)
> > +    {
> > +        char *tmp = (char *)debugstr_w(in);
> > +        tmp += 2;
> > +        tmp[strlen(tmp)-1] = '\0';
> > +        return tmp;
> > +    }
> > +    else
> > +        return szNull;
> >  }
>
> This should never happen. Places that can get a null pointer should use
> the standard debugstr_w.

oleaut32: do not crash in logging if string is NULL

[Aric Stewart]

Aric Stewart wrote:
> Meaning the place where I am finding this is different bug?
>
> In deserialize_param in the VT_BSTR case if the byte_length of the BSTR 
> that we are deserializing is -1  then BSTR_UserUnmarshal places NULL 
> into *arg.
>
> I will admit I know very little about this code but having it crash when 
> trying to log seemed very wrong.
>
> Should I change the
> if (debugout) TRACE_(olerelay)("%s",relaystr(*(BSTR *)arg));
> to a simple debugstr_w or is this revealing a deeper problem somewhere?
>
> thanks,
> -aric

By the way I am seeing this when looking at Print Preview in IE7.

-aric

oleaut32: do not crash in logging if string is NULL

[Alexandre Julliard]

Aric Stewart <aric(a)codeweavers.com> writes:
> Meaning the place where I am finding this is different bug?
>
> In deserialize_param in the VT_BSTR case if the byte_length of the
> BSTR that we are deserializing is -1  then BSTR_UserUnmarshal places
> NULL into *arg.
>
> I will admit I know very little about this code but having it crash
> when trying to log seemed very wrong.
>
> Should I change the
> if (debugout) TRACE_(olerelay)("%s",relaystr(*(BSTR *)arg));
> to a simple debugstr_w or is this revealing a deeper problem somewhere?

It's an arbitrary string so yes, it should be using debugstr_w.

-- 
Alexandre Julliard
julliard(a)winehq.org