[BUG] Applications breaking. #1400
Comments
|
Source fixes in #1401 |
|
On passive yaw--I thought the difference in yaw displacement and yaw torque would be coming from the fix to the spreading function. The passive yaw case does use spreading, but it only uses one wave direction with a single spread value of |
|
The Paraview RM3 test should be failing because it uses MoorDyn too. I've updated the logic in the applications PR so that this case runs correctly. If it fails, it can be fixed with the same solution as the MoorDyn test (TBD). if it passes I'll use that to fix the MoorDyn test |
|
@RyanDavies19 We are having some issues with the MoorDyn dll for some reason. As Adam pointed out, it seems to crash MATLAB entirely. I verified that this works locally and I also found that it works fine with the previous commit of WEC-Sim/MoorDyn (meaning it seems to be related to the recent update to 2.3.7) - see PR on my fork where MoorDyn test using 2nd to last commit doesn't crash MATLAB. I am having a hard time debugging this since it doesn't crash locally, but wanted to check if you have any thoughts? Also, @akeeste the mooring matrix test seems to be failing too due to the 'SimscapeInstrumentationLogging' error. Note that the 'SimscapeInstrumentationLogging' error seems to only occur when using rapid-accelerator mode and it does not error (locally) with normal mode. |
|
Hi all, happy to see if I have ideas on what's going on here with the MoorDyn dll. This sounds similar to FloatingArrayDesign/MoorDyn#265, but unlike what we did there we can't just ignore things here. @jtgrasb so I get it clear, this crash is happening with the DLL after new commits have been made since 8d38ef47f9e105febe007a91ecdf20ccb0c287b7? Has this commit been tested with the new DLL vs the old DLL? Just trying to get a sense of the timeline of where this started breaking. My initial thought was if input files are not formatted correctly it can make MD-C seg fault without explanation (example: point ID numbers are not in order). However I don't think that is the case here because it runs locally. Do we have any MacOS or Linux tests that run MoorDyn that we could compare? |
|
Based on my testing, it does not have to do with any new commits made on the WEC-Sim Applications repository. It seems to be the new DLL that is causing the error (from last PR) because it does not crash when testing the old DLL. We don't have a Linux test for MoorDyn because we were still working on the .so file, and we don't test WEC-Sim with Mac by default but could try to test this locally. |
|
@jtgrasb, can we try this with MoorDyn 2.3.8, the most recent release? The DLL in there now is 2.3.7, and if you remember we had issues with the API interface we were using. We commented out the troublesome code: WEC-Sim/MoorDyn@9a7e439#diff-8d5fcb6be7e83fa58a6d7139ecea1039fbca895935bc81d138f62d19947cdbb3L147 But v2.8.3 has a more permanent fix to it: FloatingArrayDesign/MoorDyn#280. |
|
The discrepancy in the passive yaw regression test:
@dforbush2 Based on the dev log, it looks like the discrepancy is coming from #1332, I believe from these lines. To handle directional wrapping better, one directional extrema (-180) is appended to the list of directions. This point needed to be extrapolated, so we switched to the spline method in Perhaps instead of adding the duplicate direction (-180 degrees) first and needing to extrapolate, we could linearly interpolate all the directional data and then append the extra direction. This would cover both the wrapping of angles and allow us to use the linear interpolation method |
|
@RyanDavies19 Just tested it with 2.3.8 and looks like its still causing the same issue where it crashes MATLAB. I am going to test a few versions out to see which version of MoorDyn the issue shows up in, and for now go back to a dll version before the issue arises. |
Some application cases are breaking in v6.1.2. Given our current release is buggy, this should be a high priority to resolve. I just reran all Applications tests using both main branches and some are still failing. Results here
MOST- these regression tests have also failed previous. It's unclear why an identical case results in some slightly different outputs. We can reset the regression data or try to test against a case without rerunning any processes with optimization or randomness. We've had prior discussions on this point, I'll try to find and link to them.
Mooring/MoorDyn (and Paraview RM3) - some drastic crash, crashing MATLAB entirely. It's hard to tell what's going on from the log, but it has failed on a 2nd retry now. This runs locally with R2023b and R2024a on windows. --> this is more or less working now. MoorDyn on MATLAB via GitHub Actions is strangely inconsistent, but it runs locally and there's a few cases of it running fine on the Paraview case or jtgrasb/WEC-Sim_Applications
Morison element - @dforbush2 @jtgrasb The full directional wave spectra Full dir spectra (resolved conflicts) #1332 seemed to break other tests running with wave spreading due to the change here. Any idea what's happening? Is there a test case or application for a full directional wave spectra case?
Multiple wave headings - failing while processing a bug signal. @akeeste will investigate --> This was related to a previous bugfix in main that got lost. Reimplemented in Bug fixes for applications #1401
OWC keeps getting cancelled and failing. --> The run time is about 35 minutes, the full test time including set-up/teardown is over 45 minutes, which is over the job timeout limit. We can increase the job timeout or remove MCR runs from the OWC test. I prefer and have implemented the latter in Application fixes WEC-Sim_Applications#80
Passive Yaw - the regression test is failing due to differences in results. --> My guess is that this related to the fixes for wave spreading in Bad bemio fcn fix #1289 (also see issues WEC-Sim v6.0 is not equivalent to v6.1.1 for OSWEC (due to wave spread bug fix) #1367 Add regression tests for OSWEC #1377). If this is the case and the current simulations are more accurate, we need to update the regression data being tested against on the Applications repo. See thread below
The text was updated successfully, but these errors were encountered: