Blender Foundation: Blender 2.x BF release: Detail: 24862 Fluid Simulator issues

Log In
New Account

[#24862] Fluid Simulator issues

Date:
2010-11-22 15:53 Priority:
3

State:
Closed

Submitted by:
gimble gimble (gimble) Assigned to:
Janne Karhu (jhk)

Category:
None Status:
Fixed / Closed

Relates to: Duplicates:

Patches:

Summary:
Fluid Simulator issues

Detailed description

Linux Ubuntu 64
Blender 2.55 Beta r33196

1. "Reverse Frames" feature in fluid domain panel appears to be non-functional:

http://users.xplornet.com/~gimble/test_249_off.avi
http://users.xplornet.com/~gimble/test_249_on.avi
http://users.xplornet.com/~gimble/test_255_off.avi
http://users.xplornet.com/~gimble/test_255_on.avi

______________________________________________________________________________

2. Fluid object fails to animate when using simple location ipo keyframes?
In 2.49 and 2.55 'export animated mesh' tool tip, it states that this option is not required in these cases.

It does the same in 2.49.
Maybe I'm misunderstanding the purpose of the 'export animated mesh' checkbox for Fluid; but when compared to what it does for Inflow and Obstacle objects, it appears to be broken for Fluid objects.

Sorry, I know you said not to mention things that were the same as 2.49, but I couldn't resist... :)
______________________________________________________________________________

3. No Drops or Float particle types any more?
If these have been "replaced / usurped" by SPH particles, how do you have
the particles move with the fluid?

4. No "Fac -Time" animation variable for fluid.
I mentioned this to Matt Ebb, and he stated that there were plans to come
up with a more universal/global solution for all physics. Any update?

5. Scrubbing the timeline while the fluid is baking causes blender to crash (not always, but often).
(Seems to occur most often if you scrub to a frame beyond what has been baked).
Fluidsim: error in reading data from file.
Segmentation fault

Please don't disable. :)

Followup

Message

Date: 2010-11-22 16:39
Sender: Ton Roosendaal
Assigning to Janne.

Date: 2010-11-22 20:24
Sender: Thomas Dinges
Can confirm case 5. Dont tested the other ones.

Date: 2010-11-23 17:33
Sender: gimble gimble
Please disregard #3 -> user error.
Sorry, I should have looked harder.

Date: 2010-11-24 11:58
Sender: Janne Karhu
1. The old functionality was lost somewhere in the conversion to using the job system, but I restored the functionality pretty much (r33278). The only change is that now the end frame (new start frame) is figured out when the "reverse frames" option is checked, so if there are changes made after that you need to uncheck and re-check the option for the changes to take effect.

2. A "fluid" type fluid object is the simplest of the different options and it can't be animated. What it basically does is takes the object and converts it fully into a fluid on frame 1. The object doesn't exist for the fluidsim after that, only the fluid that was created from the object. If you want to animate the object you have to use the "inflow" fluid object.

4. This hasn't been worked out yet, and I don't have any info on this universal/global solution Matt was talking about.

5. I didn't get it to crash, but I added some code to the place where the error probably happened just in case, so please re-test with a build after revision 33278 and report back here.

I'll keep this report still open so you have time to check if 5. was fixed by my commit.

Date: 2010-12-01 16:23
Sender: gimble gimble
Hi Jahka

Thanks for fixing the reverse frames option. A quick test:
http://users.xplornet.com/~gimble/test.avi

With regards to #5, I still get the odd crash, the only thing
that it displays in the terminal is "Segmentation Fault".

I have been trying to isolate what is causing it, but still
am not quite sure of the reason.

It seems to happen when, for some strange reason, when I set
the path for the domain, and it seems to re-create the path
underneath the defined path. It doesn't seem to do it very
often, but when it does, scrubbing the timeline seems to
result in a crash.

Hard to explain, so here's some images:

http://users.xplornet.com/~gimble/blender_fluid_panel_01.png
http://users.xplornet.com/~gimble/blender_fluid_panel_02.png
http://users.xplornet.com/~gimble/blender_fluid_file_browser.png

If you compare 01 vs 02.png, in one case it seems to use "//",
while in case 2 it enters the entire path.
In both cases I use the little folder icon to browse to the folder,
(in some cases I create a new folder, in some cases I use an
existing folder - didn't seem to matter concerning the occurance
of a crash); then I enter a file name prefix: usually "test_"
and press [Enter].

When the path name appears in the text box preceded by "//", in
most cases the fluid .bobj files are stored in the defined path,
but in some rare cases the path is duplicated and the fluid files
are stored in the duplicated path location. This seems to be the
common factor which results in a crash when scrubbing the time line
while the fluid is baking.

Here's a blend file which always crashes on my system
(especially when you set the scene Z gravity to a low value like -0.1):

http://users.xplornet.com/~gimble/test_01.blend

Date: 2010-12-02 15:38
Sender: Janne Karhu
The problem with the paths is that in some cases you have probably selected the fluid file path (with relative paths on) before actually saving the blend. Then you have saved the blend before actually starting the bake, so as a result the "relative path" is now totally wrong. I'll see if this can be fixed somehow in the future, but for now you can avoid this problem by always unchecking "relative path" when selecting the fluid file path.

Closing the report.

Attached Files:

No Files Currently Attached

Changes:

Field	Old Value	Date	By

Resolution	Investigate	2010-12-02 15:38	jhk
close_date	2010-12-02 15:38	2010-12-02 15:38	jhk
status_id	Open	2010-12-02 15:38	jhk
Resolution	None	2010-11-22 16:39	ton
assigned_to	none	2010-11-22 16:39	ton