From: Tim M. <ti...@re...> - 2007-10-14 00:22:13
|
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Melchior FRANZ wrote: > * Tim Moore -- Saturday 13 October 2007: >> If I had received only objections, I wouldn't have checked it in. > > I and AJ objected. Andy suggested an alternative (trying .osg and > .ac on objects with no explicit extension), which I don't exactly > consider an agreement. Nobody agreed, but Vivian and Jon tested > your patch. Jon reported no speed improvement. (This is not from > memory, but from the IRC log.) Jon may have tested the model in isolation, but I don't believe he tested the accompanying code changes, which are not limited to just checking for the alternate model. I've changed the model loading code to limit the amount of copying that is done for each instance of a model. Andy's suggestion doesn't work for my purposes; I don't know of a way to change all the references to generic_pylon_01.ac, etc. embedded in .stg files without rebuilding the scenery. Furthermore, this is not so much for my benefit, because I can't recreate the stuttering problems that some have reported, but for those who are having problems; I don't see a practical way to test different model files for performance differences other than with a substitution scheme like I implemented. You did suggest a property to enable / disable this feature; would that make you happier? Or an alternative search path? > > It's not that I can't deal with getting overruled. But it's IMHO > not acceptable to ignore all objections. The way to go would > have been to discuss the problem again here on the list, *before* > just committing. Either it can be resolved, or better solutions > are found. And if nobody cares to comment, you can still set a > deadline of a few days after which you would commit. I believe what I did is well within the norms of working with a development tree, including my interpretation of the discussion over the last few days. Ideally I'd check in what I did on a branch, or publish the change in my own repository, but the version control system we're using doesn't make that practical at all. The point is to get the people with the problems to try this out. If there is no difference, or the alternate model loading is breaking things, guess what? I'll revert the change! Tim -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (GNU/Linux) Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org iD8DBQFHEJ4leDhWHdXrDRURAgz6AKCYu34M5hyWAktLEjiCVk0aUGRBuQCgqaSz FZ9aKwBNd3QWVQY63pkYcq4= =aaiz -----END PGP SIGNATURE----- |