[Gnash-dev] Re: Getting window.location from plugins

[Braden McDaniel]

On Sat, 2006-09-23 at 22:27 +0200, strk wrote:
> On Sat, Sep 23, 2006 at 02:35:18PM -0400, Braden McDaniel wrote:
> 
> > Er... Meaning you got a linker error? Either something's missing in your
> > build setup (i.e., glib), or something's amiss with your installation.
> 
> Yes, sorry for bothering you with this, it's likely that I'm linking
> against the 1.2 version instead (will check).
> 
> > I haven't tried to use g_io_channel_new_file. I used
> > g_io_channel_unix_new because using pipes seemed like the appropriate
> > way to accomplish what I wanted. While I suppose you could do it by
> > writing to/reading from a file, I'd expect it to be less efficient and
> > more trouble.
> 
> Sure, but the dox say it won't be portable, and we want gnash
> to be portable.

Then use something along the lines of #ifdef _WIN32. Or perhaps NSPR has
a more portable solution; I don't know. Boost has a shared memory
library now; and I might change OpenVRML to use that in the future.

Cross-platform IPC is not a walk in the park. There are various ways to
skin the cat; and invariably they involve either adding more
dependencies or writing platform-specific code.

>  BTW, I realized we would still be much simpler
> to know the URL in the standalone executable, due to the fact
> that we'll also need to implement an internal loader (yes, could
> be done with the stream-based NP interface and proxying all
> the traffic to the pipe, but starts to sound pretty complex).

OpenVRML already does all of this. It is not trivial. But doing the
Right Thing is not always simple. By not using NPN_GetURL, you're not
taking advantage of the Web browser's cache and various URI scheme
support.

-- 
Braden McDaniel                           e-mail: <address@hidden>
<http://endoframe.com>                    Jabber: <address@hidden>