Learning Forum

[PhilippeB]

Hello,

I like Prepost but its repeated crashing when building a model is annoying (it becomes unresponsive which requires killing it to be more accurate).

It mostly occurs when changing the nodes or elements (deleting, generating though elgen or remesh, replacing nodes, splitting, etc.) but never when changing coordinates (e.g. transforming). This happens with different models that open without warning in prepost (mostly R11 and R9).

This has been happening for many years (I have been using prepost regularly but not full time for over 15 years).

After the crash, repeating the same operation on a freshly opened file usually works. If not, doing the same operation in a different version works.

So I typically save quite often a proj file as backup (quick to update) and restart from there after the crash.

But today, my proj file got corrupted somehow as I was working in 4.12 2025/R1. I cannot open it anymore in any version and I just lost two hours of tedious work. This is a new one.

I have automatic proj save (every 8 minutes) but I guess it does not keep the versions after the file is saved (why by the way?). Bummer.

So, my questions are the following:

• any possibility to extract contents of a corrupted proj file ?
• any possibility to keep intermediate .proj files? (the software should check that the file is not corrupted before deleting by the way)
• any possibility to generate crash dumps that could then be used by developpers for debug (difficult to post a bug for something that does not happen on a freshly opened file)
• when it becomes unresponsive (e.g. on element delete), what happens internally? If updating the nodes/elements database, any function to check its consistency (to prevent crashing) ? How long to wait to see if it can recover (lets say for a model <50k elements on a modern PC)?
• any other suggestion on settings/practices? 

I alternate between Prepost 4.7, 4.8, 4.10, 4.11, 4.12 (each has its own lsppconf) including double precision versions on 4.11/4.12 (seems useful with models with large ids), on two windows 10 machines with 32GB memory or more (one with integrated graphics, one with a Nvidia GPU: no difference by the way)  and the current crashes are on a small model (<30k elements, <50 parts/mat/sections, a few contacts, spotwelds and CNRB options).

Thanks

Philippe

[Reno Genest]

Hello Philippe,

Have you tried the following in LS-PrePost?

• Update your graphics card drivers?
• Use the normal renderer in LS-PrePost by pressing CTRL + L:

• Output the LS-PrePost_info.txt file by pressing CTRL + O in LS-PrePost. This file contains information about your graphics hardware, drivers and LS-PrePost settings. Note that Ansys employees cannot download files on the Ansys Forum (for legal reasons), but you can post the content of the LS-PrePost_info.txt file in a post if you want. This will help us and developers to troubleshoot the problem.

Also, note that I never save project file .proj when working with LS-PrePost; I only save the .k file and I don't see crashes often. I see crashes sometimes with very large models, but not with models with less than 50k elements.

If you are a commercial customer, you can create a support case on the Ansys Customer Support Space (ACSS):

customer.ansys.com

This way, we will be able to exchange files and better help you.

 

Let me know if this helps.

 

Reno.

 

 

[Reno Genest]

Hello Philippe,

I contacted  one of the LS-PrePost developers with your questions and here are his answers:

"

· any possibility to extract contents of a corrupted proj file ?

             LSPP doesn’t expose any functionality to our customers to handle corrupted .proj file, and therefore our customers can do nothing for the corrupt .proj file. Ideally, they should send the corrupt .proj file to us, and then we are able to debug what is going on.

· any possibility to keep intermediate .proj files? (the software should check that the file is not corrupted before deleting by the way)

              LSPP doesn’t have ability to check whether the saved .proj file is corrupted or not,  unless it is read into LSPP.

· any possibility to generate crash dumps that could then be used by developpers for debug (difficult to post a bug for something that does not happen on a freshly opened file)

               The crash dumps might be deceptive, as the real cause of crash might result from the memory corruption. Ideally, we need to get the “model + .cfile” from our customers to debug this problem.

· when it becomes unresponsive (e.g. on element delete), what happens internally? If updating the nodes/elements database, any function to check its consistency (to prevent crashing) ? How long to wait to see if it can recover (lets say for a model <50k elements on a modern PC)?

            When it comes to a model with just <50k elements, LSPP should be able to handle the element deletion without any wait. When LSPP becomes unresponsible, it probably goes into a dead loop or something else, it is hard to say how long one will wait for.

"

 

So, it would be best  if you could create a support  case and send us your model.

 

Let me know if this helps or not.

 

Reno.

[PhilippeB]

Hello,

thank you for your responses.

• Regarding the graphic card: we are talking about two different machines (one with iris xe, one with with Nvidia RTX A2000 12Gb). Drivers are automatically updated (managed by my organization) unless there are specific requirements. Are there any?
• Normal/Fast renderer: I tried it and the normal seems really slow for larger models (we usually work with models between 2 and 3 millions elements). I haven't tried it long enough to see about crashing
• Crtl-O: didn't know about it. I'll put it in the case if I open one
• Responses on the project file: ok. too bad.
• Opening a case: I'll see if I can find something with a small model that can be reproduced on at least two machines. Otherwise I guess it is pointless.

Thanks

Philippe

[Reno Genest]

Hello Philippe,

Sounds good. For the graphics card drivers, just updating the drivers should help. If your organization does it then you should be good. But, it might good to double check.

 

Reno.

[PhilippeB]

Hello Reno,

thanks for your suggestion but my drivers are not that old and if the software has been crashing for years in versions 4.7 to 4.12 under different drivers, I am not going to risk anything by forcing bleeding edge drivers unless it is written somewhere that a specific driver version solves a specific problem. 

From my point of view, an issue is the apparent randomness of the crash or the dependency to the system which may prevent the debugging. I saved model and cfile that crashed prepost 4.11 and 4.7 yesterday on a machine, but it works fine on another machine today. Typical.

It is pointless to send I guess, suggesting that these crashes will not be solved based on the cfile only and that some other form of information is needed to improve the quality of the software. The developers may say that it is due to specificities of the system but that is the name of the game for all software and many software packages are stable. It's too bad because prepost really does a lot of things.

I'll stop here. 

Thanks

Philippe

[Reno Genest]

Hello Philippe,

The best is to create a support case, send us your model, and specify the steps to reproduce the crash. We will try to reproduce the problem on our machines and if successful we will file a bug and the developer will have a look. You can create a support case here:

customer.ansys.com

Have a great day!

 

Reno.