A challange! Test the graphics card/ driver issue for yourself!

[Kricket]

The problem is the crashes aren't 100% repeatable (at least for me). If they were it'd be nice and easy and we send them off to support@hash.

One thing that usually gets me is Creating a new percentage pose. Did it right after I read your post. Created a new pose, tweaked some CPs, closed the window. CRASH.

Didnt change hardware accel yet, wanted to see if I could keep doing it. Opened AM back up, new pose.. closed window and nothing. tried again and again, AM still worked. Created Models, Actions, Chors and still it ran.

Went to tools > options and changed to Direct3D and before I was able to even hit OK it crashed.

But when I finally did set HW Acceleration to None I wasnt able to make it crash after 5 min of trying.

Running a ATI AIW 9700 on Win2k. Latest drivers.

[phil3d]

Sounds like a driver issue to me.

Go to http://www.ati.com/support/products/radeonwin2kpreviousdrivers.html
and download older drivers until you get one that works.

The one that works for me is
CATALYST™ 3.2
Driver Build 6.14.01.6307
7-84-030228a1-008040c

It's the last of the Catalyst 3.3 series

It's from MARCH of this year, and is the last one that tests ok on my machine.

Good luck! and thanks for doing the test. If you do try the older driver and it works, please post again about your success.

Phil

[Hookflash]

phil3d:

I've been getting pretty good stability with 10.5h. I just tried both yours and Krickets problem scenarios, and neither one resulted in a crash. I'm using an ATI All-in-Wonder 9700 Pro with Catalyst 3.7. I run at 1600x1200x32 with default OpenGL and DirectX settings, and I use OpenGL with A:M.

[phil3d]

Originally posted by Hookflash
phil3d:

I've been getting pretty good stability with 10.5h. I just tried both yours and Krickets problem scenarios, and neither one resulted in a crash. I'm using an ATI All-in-Wonder 9700 Pro with Catalyst 3.7. I run at 1600x1200x32 with default OpenGL and DirectX settings, and I use OpenGL with A:M.
That's great! What OS and service pack are you using? - and is there anything else you think helps make your AM install stable?

Thanks for getting involved!

Phil

[Monty]

Please try this.
1) Open up A:M.
2) Open the library panel
3) Right click to create a new folder
4) While the folder name is still selected, right click and select delete.

Brings me to the desktop everytime.

3DLabs Wildcat VP870, driver version 1.2.1
Win XP Pro SP1

Dual Xeon 2.8GH, 1 Gig Ram

[phil3d]

Originally posted by Monty

1) Open up A:M.
2) Open the library panel
3) Right click to create a new folder
4) While the folder name is still selected, right click and select delete.

When you provide these reports, please make sure that you have followed ALL the steps from the first post of this thread. Since I did not know if you had, I ran the test myself with the results being the same with hardware accelleration on or off.

So, in this case I do believe that you have indeed found a bug in the program, if you think something so unlikly to be an issue for most folks is worth getting sorted out, feel free to send a proper bug report to support@hash.com

Folks, please remember that the purpose of this thread is to define and deal with WINDOWING issues - cases where the hardware accelleration provided by your graphics card causes serious Interface issues that are distructive to your creativity. I don't think I can give a "phil3d's wrong" award in a case like this. :rolleyes:

[Monty]

Hey phil3d dont get jumpy! I dont know if this is abug or an issue with the graphics card so I posted. I dont want your award anyway. Thanks.

[Monty]

BTW, I didnt get any message saying that Master.exe caused a fault or something (like what Martin said in his e-mail to Wegg), so my question is - Is this a Video Card issue?

[hoochoochoochoo]

So, in this case I do believe that you have indeed found a bug in the program, if you think something so unlikly to be an issue for most folks is worth getting sorted out, feel free to send a proper bug report to..
:eek:
do you think anyone else will respond if you continue to reply like this?
Could have been said in a better way Philip..

[PJC]

Originally posted by Monty
Please try this.
1) Open up A:M.
2) Open the library panel
3) Right click to create a new folder
4) While the folder name is still selected, right click and select delete.

Brings me to the desktop everytime.

3DLabs Wildcat VP870, driver version 1.2.1
Win XP Pro SP1

Dual Xeon 2.8GH, 1 Gig Ram

yup, same here. did you report it yet?

- pjc

[odinseye2k]

Seems to be a common prob for me as well... AM doesn't seem to clean itself up the best a lot of times -

This applies to assumed deleted spring systems, splines (which may explain why models corrupt and you can never quite work the same without doing the cleaning up in the file yourself - basically it clobbers its own data in the model file).

Sent the bug to Steve a couple of versions ago. Assuming they're still working on it since it would be a fairly involved thing to sort out.

[phil3d]

Originally posted by Monty
BTW, I didnt get any message saying that Master.exe caused a fault or something (like what Martin said in his e-mail to Wegg), so my question is - Is this a Video Card issue?
I doubt it, as it still happens when videocard accelleration is turned off. In what you discribe you have generated a situation where the application is still trying to do something (input a char string value) when you tell it to do a second thing (delete the folder discription from the library file) at the same time, I would tend to belive that this is an application error, and should be reported.

I apologise for responding the way I did, but I felt that this is more of a "well-don't-do-that-again" type thing, then the kind of issue that has lead to the rabid-anti-Hash trashing thats prevading the other active threads. I started this thread to try to get information that can help FIND A SOLUTION for everyone. I DO appriciate anyone who goes to the trouble of doing the process outlined in this threads first post. "Me too's" that follow the same procedure are fine too. If this thread degenerates into a "chat" with the majority of it's input not directly related to the first post, it will be very difficult to refine the input into usable change information.

[phil3d]

Originally posted by PJC
yup, same here. did you report it yet?

- pjc
PJC is that with and without hardware acceleation enabled?

[odinseye2k]

1.4 GHz Athlon XP
512 RAM
Abit Mobo

Abit Siluro GeForce 3
OpenGL original drivers from Abit
Windows XP

Had three and four windows open, all in shaded w/ wireframe, manipulated like crazy - no crashing (although a mite slow with the complex model I had up).

Love that Abit - even the infamous WinDoze 98 liked this system.

[PJC]

Originally posted by phil3d
PJC is that with and without hardware acceleation enabled?

both. DX and OpenGL

-pjc

Quadro4 380 XGL/AGP/SSE2

[Ran13]

I wish my crashes were as easy to duplicate, phil.

But they aren't.

Sometimes it tanks while stitching a new spline, sometimes just deleting a CP from the end of an unconnected spline introduces me to my desktop. Sometimes I can hit Undo 3-4 times without a crash, sometimes one Ctrl-Z will do it. I realize that Undo has always been a bit touchy (an understantement if I ever uttered one), and trying to undo a complex procedure is asking for trouble, but I got an instant undo crash last night trying to undo a simple translation of a CP. This is something that v8.5 could easily undo several times over without crashing.

I installed the 10.5h patch and was able to work for almost an hour last night before deleting an unconnected CP sent me to the desktop. So that's a bit better...I guess. :p

Here's another little anomoly I ran into last nite:

http://users.choice.net/~ran13/step1.jpg
In v8.5 you could select a single CP, then select the whole central spline, and the Properties window would maintain the pivot settings for the original selected CP. You could then manually enter a X-axis transform offset based on the originally selected CP's pivot position on the X axis to bring the entire central spline "seam" back onto the Y axis (X=0).

In this first pic you can see that my central seam is slightly off the Y axis. The Properties windows states that the pivot of the single selected CP is at +0.06 on the X axis.

So I enter "-0.06" in the X axis transform field in the Poperties panel, like so:
http://users.choice.net/~ran13/step2.jpg

Then I hit enter, and this is what I get:
http://users.choice.net/~ran13/step3.jpg

The pivot location data in the Properties panels does not immediately change, but only updates when I click on it. Upon clicking on it, it shows that the CP is now @ -0.09 on the X-axis. So transforming the CP by -0.06 on the X-axis from 0.06, moves the CP to -0.09 !!?!?!?!?

What kinda Chinese math is that?

And I'm quite certain this has absolutely nothing to do with video drivers.

When a 3D app can't do simple addition and subtraction of floating point numbers, what does that tell you about what other lunacy is going on under the hood?

[phil3d]

Wrong tool for the job. Click the "show manipulator properties" button.

[Ran13]

That's bullsh*t phil and you know it.

If the data wasn't meant to be manipulated this way then why are the fields editable? In works fine in 8.5!

The manipulator tool, without the little "Manipulator properties" widget won't let specify exactly "0". And that manipulator properties widget is a pain in the a** to model around.

I WAS able to get it to move to "0" on the X-axis using it, but trying to undo the simple translation goes like this:

1st undo: nothing happens
2nd undo: properties panel dissappears
3rd undo: HELLO DESKTOP!!

In addition to that, despite the properties panel stating that the pivot is @ 0.06 on the X, the manipulator properties widget says it's at "0.02"?!??!?. Changing that to 0 does move it to the proper place, and only then does the poperties panel say the pivot is at 0 on the X.

If the app can't report it's own data correctly to itself, how can anyone expect it NOT to crash incessantly?

I've worked in 8.5 a lot..I mean a LOT. And as someone who has gone thru function curves with a fine tooth comb on complex animations too numerous to count, I've found that, in A:M, "0" does not always mean "0". If you've been knee deep in function curve editing you know what I'm talking about. I've popped open f-curves that look like the slope of the Matterhorn only to find that the only 2 keys set were both at 0 on the first and last frames of a cycling animation, but apparently one "0" was bigger than the other "0" causing the slope.?!?! It took manually entering "0" as the keyed value for both keys to get the f-curve to go flat and get rid of the "jitters" in the keyed bone when the animation cycled.

[phil3d]

Originally posted by Ran13
That's bullsh*t phil and you know it.

Well dam! its a good thing your available to tell me how I've been using AM wrong all this time. :rolleyes:

The whole process of using this tool correctly has been gone over ad naseum on the AM users list. Here's a link to the archives.
http://venice.killernuts.org/animaster/
I suggest you take some time to catch up with the AM users who have been using versions 9 to 10.5 where this is done in a different manner then 8.5.

BTW, I have no need of help in sorting out AM issues as they come along, I know how to work with and around AM issues as they come up to enable the delivery of product for my clients. The goal of this thread was to help those who do need help, and to provide Hash and the GPU makers with the info they need to make AM stable for everyone.

Is this somehow against you intrests? I only ask because you seem to be going to great lengths to sabotage ANY positive contributions on this and the other active AM threads.

Phillip Leavens

[Monty]

Okay, sent that note to Steve about the bug.

Hey, got an error message - user Steve@animationmaster.com unknown. Is this e-mail address correct?

[Kricket]

try support@hash.com

[Monty]

Thanks Kricket. Just sent the report.

[phil3d]

Originally posted by Monty
Okay, sent that note to Steve about the bug.

Hey, got an error message - user Steve@animationmaster.com unknown. Is this e-mail address correct? Sorry, my bad. I've corrected it now.

Phil

[Monty]

Hey, got a reply from support. Known issue - working on it.

[Ran13]

If the app can't report it's own data correctly to itself, how can anyone expect it NOT to crash incessantly?

I wanted to say it again in case you missed it.

Gee...I thought the Properties Panel was there for a reason. And that it was editable for a reason. I guess it's just another one of Hash's cosmetic "features" that don't live up to it's billing.

And if the Properties panel data is in error, how can anyone be certain the data sent down the pipe to the vid card is correct...etc, etc.

No, I'm not here to simply "Bash Hash"...I just want to make it abundantly clear that the current version of this app is fundamentally flawed and attempts to correct it's faults by looking at the video card drivers is nothing more than cow-towing to Hash's long history of blaming everyone but themselves for the instability of their app.

Good luck in your attempts, but I wouldn't hold my breathe.

[Ran13]

Just for kicks, I did a search on the maling list archive for "Properties Panel" in the content and here's some of what I came up with:

2000-05-27 07:52:42 justin@jb-av.com
Re: Aligning CPs?


To: animaster@hash.com

From the key tappings of Dennis Borruso:

> Does anyone know of a way to align a group of selected CPs? Whenever I
> am doing a copy,flip,attach, I end up aligning the center CPs one by one to
>align at 0 on the x axis. It takes a while. Is there a short cut for this?

- Group all the CPs you want to align
- In the Group tab of the Properties panel, scale them all to 0 on the X
axis (or whatever axis you want to align them to). This will give them all
the same X value in case they don't match already.
- On the Pivot tab, look at the X position. For example, it may read 0.81
- Back in the Group tab, enter the opposite of that value into the X offset
field. Using the above example, you'd enter -0.81, then hit Enter.



Justin Barrett
Character / Animator
http://www.jb-av.com/



2001-05-22 20:51:05 mattra@ix.netcom.com
Re: How to move object/model to precise coordinates?

To: <animaster@animationmaster.com>

> 1 - How can one translate an object to a specified point?
>

Click a spline section so you only select one point (don't use the group
tool at this stage), go to the selected cp tab of the properties panel,
you'll have x,y,z coordinates of that point, now invoke the group tool and
select the rest of the points you wish to move, then type in the x,y,z
coordinates you want your initially selected cp to move to, the rest of the
points will follow suit.

> 2 - Is there a way to move an object based on its center of gravity?
> Example, move a ball or circle to center on a specified x,y & z point
in
> the coordinate system?

Just select all the points in your ball, go to the pivot tab in the
properties panel, the pivot location will reflect the model's offsets from
the origin, so just copy the pivot numbers into the group offset fields for
each axis but make the value opposite, plus to minus and vice versa. When
you complete the operation the ball will be centered, check the pivot tab
and it should read 0,0,0. The pivot tab is very important for exact
placement modelling.

Matt Andersen


I saw very little on the Manipulator Properties widget, only a few posts from people who couldn't find the button in the first place. Apparently at one point in the v9 debacle, the installer failed to put the button on the toolbar. :p However, I saw nothing of any "innaccuracies" noted in the properties panel or any hint that the properties panel was the "wrong" place to manipulate data. Only the above posts that proposed a method exactly like I was trying to accomplish.

So much for the "wrong" tool theory, phil.

[phil3d]

I've tried to duplicate your problem in 10.5h multiple times now with no success. I used the properties of an un-named group to locate the pivot of a cp, and then tried to "0" the cp's position on the "x" axess by inserting the reverse value of the cp's pivot in to the "x" translate field. In everycase it worked fine.

Here's why I say your using the wrong tool. If you are trying to manipulate a cp directly, why would you go to the trouble of defining its position fom the pivot? The Show Manipulators property tool DIRECTLY provides a way to know and change the position of a CP. If the cp's "x" value is 12 and you want it to be "0" just change the 12 to 0 and your done, that's WHY the tool exists. It can be moved to get it out of the way and is FAR quicker to access the properties needed for translation, scale, and rotation of ANY given cp or group of cp's this way.

Try it for a while, there are a lot of changes from 8.5 to 10.5 in the interface that at first are hard to understand. After you spend some time with them I think you'll think differently about it.

[Kevin Sanderson]

Randy, do a search by contents on Show Manipulator. You'll find many helpful posts.

[Ran13]

OK guys...I understand the Man. Props deal. I don't like it, but I can live with it. I find the Man. Props widget HIGHLY invasive, especially on a dense mesh. But that's another story... ;)

But look again at the pics!! You both seem to be missing my point. I don't know why you can't duplicate this, phil. It was very easy to demonstrate for the screen grabs I did.]

Step-by-step:
1) Open existing project.
2) Select View/Show Properties.
3) Select a CP.
4) D-Click on the X-axis translate field in the Properties Panel to access it's edit box.
5) Enter a value.
6) Hit Enter
7) CP moves to a new position totally irrelevant to the entered data!

I'm REALLY NOT making this up!!! And if you think I am, check the pics closely, you'll see that I started with a freshly opened project (no asterisk) and that I only changed data on the X axis (the asterisks next to the fields indicate where the changed data is...and I didn't change squat on the Y axis, and there is no asterisk there, but the data is different from pic 1 to pic 3!)

Look at the very first pic. The CP I have selected is very close to the -2" mark on the vertical ruler, yet the properties panel reports the pivot of the single CP @ -5.27"!!!

When I used the prop panel translate fields to move the CP on the X-axis ONLY, not only did the CP jump .09" further than requested, but it also dropped to -5.28" from -5.27" on the Y-axis according to the prop panel.

I just want to say, again, that I realize this is not the "preferred" method in v9/10, and these values are miniscule, but the erroneous data in the prop panel still indicates that there IS erroneous data being passed to it from somewhere. (And it ain't from the vid card river, phil ;) )

Doesn't this type of thing throw up a red flag for anyone but me?

[phil3d]

Ran13, I understand that this seems like a real issue to you, and in your shoes, just coming from 8.5, I would probably have the same concerns.

I don't have 8.5 installed any more, but if I remember correctly there was a tab in the properties group for the selected cp. You could get values in it by clicking or shift-clicking on a cp. At that point the values would be editable to translate etc your cp. These values reflected the cp's position in relationship to the world space. (for instence, if the cp is located at 123,89,235 the values will reflect that.

There was also a group tab for group translation that had the values at "0" (reflecting local space)

The pivot information also provided a refrence based on world space, so it COULD be used to infer the position of a cp and then move it using the local space reference. I doubt that is how you worked in 8.5 though (at least I hope not - it's alot of work!)

Translating, rotating or scaling a cp (or group of cp's) is easiest to do when you have the World Space values to work from. The cp is located 12, 15, 9 and to move it to 3,12,89 you just type in the values.

Now the difference in 8.5 and 9 and up.
Since translating, rotating or scaling a a group or cp is done so frequently when working with AM, the programmers at Hash decided to move the "cp position tab" that was in 8.5, to a onscreen widget that makes it accessable even with the properties panel closed. Once I got used to it I discouvered that I love it, you can model full screen, and everything you need (for splineing) is quickly available.

Now there are some definite application issues with values displayed in some fields. I KNOW that e1998203 is NOT a valid coordinate. It could be that you are getting some kind of error from whatever is at the root of these issues, and the next update will hopfully "fix" it. Whatever, use the widgit thats what it's there for.

(He, he... remember.. "Luke, use the instruments, that's what their there for") Sorry, sillly moment.

Phil

[Ran13]

I doubt that is how you worked in 8.5 though (at least I hope not - it's alot of work!)

Most commonly for centering the middle seam before copy-flip-attach, but on other somewhat rare occasions as well...usually when I want all the splines in a selected set to lie on exactly the same plane...like the bottoms of a character feet. You could windows-select the central CP's, scale to "0" on the plane's perpendicular axis using the Group tab, and Shift-Click on any one CP on the central spline, and use the pivot data to translate the whole spline to 0 on the appropriate axis.

I played with it this evening and found that the same procedure works with the Manipulator properties panel. I hotkeyed the widget and now it's not as painful to deal with. :D

I also did some "time 'til crash" experiments with the D3D and OGL drivers. Since I'm at the point in the model where I go thru and nudge CP's around to get the splines as smooth as possible, I figured that it would be a good time to test which driver was the most stable for me. I was able to work much longer (almost twice as long) with the OpenGL driver than with the D3D driver for my particular video card (ATI Radeon 9500 Pro). So I suspect (and always did) that at least SOME of my issues are vid card related.

It also seems that "time 'til crash" is greatly reduced by lots of "Ctrl-S", which points towards issues with memory management. :hmm:

Onward thru the fog... :wip:

[phil3d]

Originally posted by Ran13
I hotkeyed the widget and now it's not as painful to deal with. :D
-------------------------------------------
I was able to work much longer (almost twice as long) with the OpenGL driver than with the D3D driver for my particular video card (ATI Radeon 9500 Pro).

Hotkeying the widget is a great idea! I hadn't even looked to see if it was possible. Thanks..

D3d is almost unusable on my Radeon 8500le, slow and with MAJOR re-draw issues. OpenGL works great though, so I don't worry about it too much, except when I want to upgrade drivers and OpenGL tanks too.

Oh, on the center seam thing, you probably already know this but "shift click" will allow you to select 1 cp of a group for setting it's position. So, to do the same, I "," select the spline, hit "s" to scale, type "0" in the widget, hit "s" again for the standard manip, shift-click a cp in the selected group (if not already selected) and enter "0" in the cp position setting of the widget. Wow, that take about 10 times longer to write then to do!!

Phil

[Ran13]

D3d is almost unusable on my Radeon 8500le, slow and with MAJOR re-draw issues. OpenGL works great though, so I don't worry about it too much, except when I want to upgrade drivers and OpenGL tanks too

Yea, I've found that ATI's OGL implementation to be good for most apps. I'll stick to OGl in A:M as well.

Tonite I'll re-set my hardware acceleration down to zilch and work with both drivers to see how long I can go without crashing. I'll post my results sometime tomorrow.

As with most software issues, I think A:M's general "wonkiness" is likely a long list of contributing factors, not the least of which is IHV video drivers. Throw multiple processors, multiple NIC's, multiple monitors, co-existent IDE & SCSI drives, tablet drivers, software firewalls, etc. (all of which exist on my system) into the mix, and it becomes a rather dense minefield of possible causes of instablility.

I firmly believe that Hash has to learn what other 3D app creators learned long ago, standardized MFC apps, especially with multiple document interfaces, are no place for high-end 3D.

[Ran13]

Well, this morning I turned hardware acceleration to none. I worked in OGL for about 2 hours. I had dumps to the desktop when deleting CP's and when connecting CP's, even with mucho Ctrl-S'ing. I got past a "Copy-Flip-Attach" without problems (other than the normal screw-ups around 5-pointers), but after applying porcelain, I went to adjust the "Normal Weight" percentage and was introduced abruptly to the desktop.

Now, when I try to open the project I get "Invalid project file...etc." I then created a new project and tried to import the model, which was saved external to the project, and get "Invalid model file...etc." :eek:

Needless to say, I'm done with A:M for a while. Could someone let me know when Mr. Hash wakes up and smells the coffee? :thumbsdow

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.