[Verse-dev] Saving and Loading

[Emil Brink]

[VML tools]
> I tried a little more complex geometry (> 12000 polygons) and the saver 
> did not manage to produce any vml, but crashed while reading the data 
> from a Verse-server.

Oops. :/ Sorry to hear that, of course. I wish there was something I
could do, like reproduce the problem locally. If possible, when you
have similiar data in a server, feel free to drop by #verse on the
Freenode.net IRC network. I can run the saver from here, and perhaps
get some debugging done.

> But it is not the amount of data that matters, I think, but the type
 > of the data.

Oh?

> My model includes a lot of material nodes, in addition to the object and
 > geometry nodes. Are there some limitations to the structures which can
 > be saved? What if the data on the server is incomplete (I think it might
 > be)? What kind of vml-files are produced then?

Good questions. There are no such limitations that I know of, but I
didn't write Enough or the saver, Eskil did. I would be very surprised
if there were problems caused by things being incomplete, though. That
is something he tends to be very aware of, since he designed it that
way. :)

> I traced the crash to calling verse_callback_update(). So obviously a 
> callback registered by "enough" somehow causes a segmentation fault.

Could also be something deeper, like a crash in the network service
code. Those are not common, however, so a callback is perhaps the
most likely culprit.

> The problem can be also in compling the "enough" library. I am not able 
> to compile quelsolaar without modifying the code. For example, 
> glutGetProcAddress is not included in GLUT 3.7. Is it possible to 
> compile just "enough" without having to compile any glut-related stuff, 
> which causes problems.

Yeah, I've been hit by that problem too, it seems GLUT is lagging on
some platforms. Perhaps one should do another target, like the GLFW
library, it sounds like it is getting more attention than GLUT is,
these days ...

> If we could get the saver work properly, it would be very helpful, 
> because the vml-files can be send by email and saved to disk. This would 
> solve the problem of creating test scenes for the audio simulation.

Absolutely, these are some of the goals.

/Emil