latest and greatest (august 31 2004)

Anything related to a beta release of JPatch: Bugs, enhancements, general discussion...

latest and greatest (august 31 2004)

Postby sascha » Tue Aug 31, 2004 10:37 am

I've uploaded a new development version to
http://www.jpatch.com/temp/jpatch_20040831.jar

(as always, start with java -jar jpatch_20040831.jar)

changelog:

31 Aug 2004
===========
* Models can now be opened without creating a new model (thus merging the new file with the current model) - this is not undoable!
* Added a color chooser to the material editor.
* Added a select button to materials - it will select all controlpoints of patches of the current material.
* Changed the XOR selection mode (CTRL modifier in default tool) to only remove points.

I'd say, that's it for version 0.3 - I'll do some tests on my own for the next few days (especially the automirror requires more testing), fix bugs I find and will eventually release it (I'm not sure about the name - JPatch 0.3 beta 2, JPatch 0.3 Release Candidate 1, JPatch 0.3... what do you think?)
If you've got time, please download this preview and report any bugs etc.

I think it's time for a new release (and it's time to get rid of that "beta" tag in the release name), so for now I'll focus on bugfixing. I'll not add any new feature for this release (feature requests are appreciated though, they'll only have to wait for version 0.4...)

Thanks a lot!
-sascha
Last edited by sascha on Thu Jan 27, 2005 12:45 pm, edited 1 time in total.
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

possible undo of model add

Postby zmemw16 » Tue Aug 31, 2004 10:15 pm

if you created a selection of all points of first model, imported next model and had an invert selection mode as opposed to hide?

you could then delete the inverted selection ??
zmemw16
 
Posts: 77
Joined: Sun Jun 06, 2004 5:35 pm
Location: united kingdom

Postby sascha » Wed Sep 01, 2004 10:10 am

Your point. I quickly hacked togehter the open (merge) option - sure, it should create a new selection for each loaded model, so it's easy to select, move or delete them.

I'll try to make it undoable too, but that may be tricky and I'm not sure if it's worth the effort, as for the long term I favor a different approach allowing to open multiple models simultanousely...

Extending selections (or creating new ones) for auto-mirrored shapes is an open issue too...
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

xor function on control key istr

Postby zmemw16 » Wed Sep 01, 2004 2:03 pm

quick rethink on above invert function

invert is effectively select the 'original model' selection, hold control and drag across the lot
hit delete

you need to make sure though of creating the selections, you might possibly want to revert to as you go

stephen
zmemw16
 
Posts: 77
Joined: Sun Jun 06, 2004 5:35 pm
Location: united kingdom

Postby sascha » Wed Sep 01, 2004 3:16 pm

OK, now I understand :-)
I'll add that too (in fact it was part of the old JPatch 0.2 version, I simply forgot to re-implement it in 0.3...)
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

null pointer exception

Postby zmemw16 » Mon Sep 06, 2004 9:53 am

os libranet 2.8.1 sun j2sdk1.4.2_04 and jpatch_20040831.jar
command line as posted elsewhere with 800Mb allocated

open a largish model
select all
with control held, unselect what i don't want
hit delete

sometimes i get a line from somewhere in remaining points to the top
left of the screen, sometimes i don't, seems to be a function of how fiddly it is unselecting points.

then doing save i get a zero sized file and the following
in the messages
jpatch.entity.ControlPoint.java 1727
jpatch.entity.Curve.java 280
jpatch.entity.Model.java 115

all seem to make sense using source files of jpatch0.3beta1

istr recall finding some points really close together doing this?


stephen
zmemw16
 
Posts: 77
Joined: Sun Jun 06, 2004 5:35 pm
Location: united kingdom

Postby sascha » Mon Sep 06, 2004 2:05 pm

there are some bugs causing JPatch to accept illegal operations (e.g. it's not checking for all possible mistakes when adding hooks).
This leads to inconsistent models which will lead to all kind of problems (model can't be saved, problems with undo/redo, problems when deleting points, etc...).

once you've got a model in some kind of "illegal" state, when it's saved there not much you can do. you can try to find the problematic sections and try to delete them, but if this fails too, there nothing that can be done (ecxept correcting the problem "by hand" in the xml file...)

I've found one problems with hooks which might cause the problems you described, I'll fix it asap. But the chance is high that there are more such problems, so use the modeler with care, save often, and if you think that something went wrond, DONT overwrite your existing files and try to reproduce the problem with a model being as simple as possible.

I'll change the way JPatch saves file to not overwrite old files but to create backups, to prevent loss of data when you run into a bug and save the (bad) model...

can you send me the problematic model, I'll try to figure out what's wrong with it

ps: help->about will show the version number and the date/time of compilation...
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria

help:about

Postby zmemw16 » Mon Sep 06, 2004 6:44 pm

shows the version of jpatch, but not the version the model file may have been started with.
as i said i tend to track your latest and greatest, and i have model files back to the very first beta after the 2.1 release, so far i've been bitten twice i think

currently using 20040831 version
i was in the process of learning from my earlier mistakes and trying to snaffle bits of a previous model, eg starboard wing, taileron and a drop tank from the previous model (read latest-1 (jpatch280704??)) and doing open(keep) into latest, that was the idea

is there any way of displaying a bad hook? silly, if you could display it as bad, you could unpopulate it

anyway of walking along a spline calculating distances and testing for points that are extremely close to another point? really silly values (say <0.5%) of some average, particularly where multiple spline are joined.
sounds distinctly time destroying :-)


any possibility of posting descriptions of the various operations that tend to lead to this 'illegal' state, i suspect the process of reducing a selection by use of control + mouse select might well be one?

stephen
zmemw16
 
Posts: 77
Joined: Sun Jun 06, 2004 5:35 pm
Location: united kingdom

Postby sascha » Tue Sep 07, 2004 11:43 am

any possibility of posting descriptions of the various operations that tend to lead to this 'illegal' state


If I find one, I usually fix it ;-) but see below...

Right now there's one open issue: You can only hook the end of a curve - but if you've got two curve ends which are already connected, it would still allow you to hook those two points - which is not clean.

Generally speaking, JPatch is beta, so don't trust it ;-)
Especially after complex operations (copy, extrude, lathe, mirror, delete/remove, undo/redo, etc.) check if the result looks feasible - and if not - don't overwrite the existing file! Try to reproduce the problem with the simplest possible topology and report it as a bug.

any possibility of posting descriptions of the various operations that tend to lead to this 'illegal' state

Theoretically all of them :-(
The problem is that most operations have to check for a number of special cases - so, just for example, an operation could work fine in most cases, but could fail when e.g a hook is involved (because this is a special case and is treated seperately, and there could be a bug in the spacial "hook handling" code).
Unfortunately this makes testing a lot harder, sice I have to test all those special case seperately (and sometimes I don't even know what's being handled seperately)...

Therefor I am working on automated testing methods, but right now they are not useable.

is there any way of displaying a bad hook? silly, if you could display it as bad, you could unpopulate it

Not so silly. I could think of a method which analyzes the model and looks for all kind of "illigal" states. It could get quite complex (i.e. slow), so I wouldn't do that per default, but I could think of a debugging switch to turn on the model-checking after each user operation, or in extreme cases even after each "atomic" action JPatch performs behind the scenes - so if you suspect an operation to mess up something, you could turn on that "debugger" to check it...
sascha
Site Admin
 
Posts: 2792
Joined: Thu May 20, 2004 9:16 am
Location: Austria


Return to Beta

Who is online

Users browsing this forum: No registered users and 1 guest

cron