OpenVSP 3.51.3 Released
8 views
Skip to first unread message
Rob McDonald
Aug 17, 2026, 5:42:43 PM (8 hours ago) Aug 17
to OpenVSP
While this version breaks from the strict 'speed only' theme of .1 and .2, this release is still pretty small and doesn't warrant a new .0 release.
It includes one fix for the only problem found so far in .1 or .2. This was an esoteric issue with the AngelScript wrappers for some OpenVSP classes. These only manifested when running scripts that used certain language features on these classes. Thanks to John Dannenhoffer for finding a failure case at the Workshop.
That lead me down an AI driven path to turn the API example code (and other sample scripts in the tree) into unit tests that can actually report failures. This sudden increase in tests consequently turned up some bugs. This also has the effect of making the documentation code samples a bit ugly. If this is a real problem, let me know and I'll see if there is a way to get clean examples while keeping the value of the tests.
As it turns out, our inertia calculation for thin-shells has been wrong forever. Since the major contribution is the parallel axis theorem part (which was right), our answers were close enough to not raise suspicion. Now they are correct.
The search and replace variable names in Advanced Link code could screw up in certain situations. It shouldn't eat your code now.
CFDMesh with a symmetry plane may be vanishingly faster. Nobody will ever notice.
There were a few other fixes in this sweep. Nothing major, but good things to fix.
After the sample code and unit testing sweep, I asked AI to audit the API for completeness. This was inspired by the observation that there was no SetContainerName() in the API. Thanks to Cooper Cook for pointing that out.
We started with a pass looking for asymmetric pairs -- Set/Get, Add/Delete, etc. After that, we searched for any GUI button that didn't have an API equivalent. For all these things, AI added the API call, registered it with AngelScript, wrote the help documentation including AS and Python examples - including test conditions making the example into a unit test. While I'm sure it isn't perfect, it should be better.
Another Workshop discovery, Max Lindstrom found that you couldn't work with matrix type spreadsheet UI's (say for Attributes or Results). This was caused by a recent change I'd made to increase pointer safety in that area. Searching for the same problem elsewhere turned up another latent bug in AttributeScreen.
As it turns out, ancient SWIG doesn't like 'using std::' directives. Since we used std:: inconsistently in the API header, seemingly random Python API calls failed to work on RHEL8. That should be sorted.
The Python MANIFEST.in should be platform agnostic for any users installing the OpenVSP package with pip.
Jaime Benitez did some debugging and enhancements for VSPAERO moving body full unsteady analysis. They should be useful for people who drop things.
Roman Prokopyshyn had help from Claude fixing some problems with one of the VSPAERO example cases that hadn't been touched since before the thick/thin era.
Features:
- API code examples now serve as unit tests
- AI Audit for API completeness
- VSPAERO Outputs more stuff for dynamic analysis
Build System:
- Update AngelScript to v2.38
- Improve Python MANIFEST.in and packaging
Bugs:
- Fix AngelScript registration of vec3d and Matrix4d
- Fix thin shell inertia calculation
- Fix many small bugs identified by new unit tests
- Fix matrix spreadsheet UI referencing
- Fix DeltaFlatPlateDragArea for unsteady cases (thanks Jaime)
- Swept_Wing_API example VSPAERO script brought current (thanks Roman)
It includes one fix for the only problem found so far in .1 or .2. This was an esoteric issue with the AngelScript wrappers for some OpenVSP classes. These only manifested when running scripts that used certain language features on these classes. Thanks to John Dannenhoffer for finding a failure case at the Workshop.
That lead me down an AI driven path to turn the API example code (and other sample scripts in the tree) into unit tests that can actually report failures. This sudden increase in tests consequently turned up some bugs. This also has the effect of making the documentation code samples a bit ugly. If this is a real problem, let me know and I'll see if there is a way to get clean examples while keeping the value of the tests.
As it turns out, our inertia calculation for thin-shells has been wrong forever. Since the major contribution is the parallel axis theorem part (which was right), our answers were close enough to not raise suspicion. Now they are correct.
The search and replace variable names in Advanced Link code could screw up in certain situations. It shouldn't eat your code now.
CFDMesh with a symmetry plane may be vanishingly faster. Nobody will ever notice.
There were a few other fixes in this sweep. Nothing major, but good things to fix.
After the sample code and unit testing sweep, I asked AI to audit the API for completeness. This was inspired by the observation that there was no SetContainerName() in the API. Thanks to Cooper Cook for pointing that out.
We started with a pass looking for asymmetric pairs -- Set/Get, Add/Delete, etc. After that, we searched for any GUI button that didn't have an API equivalent. For all these things, AI added the API call, registered it with AngelScript, wrote the help documentation including AS and Python examples - including test conditions making the example into a unit test. While I'm sure it isn't perfect, it should be better.
Another Workshop discovery, Max Lindstrom found that you couldn't work with matrix type spreadsheet UI's (say for Attributes or Results). This was caused by a recent change I'd made to increase pointer safety in that area. Searching for the same problem elsewhere turned up another latent bug in AttributeScreen.
As it turns out, ancient SWIG doesn't like 'using std::' directives. Since we used std:: inconsistently in the API header, seemingly random Python API calls failed to work on RHEL8. That should be sorted.
The Python MANIFEST.in should be platform agnostic for any users installing the OpenVSP package with pip.
Jaime Benitez did some debugging and enhancements for VSPAERO moving body full unsteady analysis. They should be useful for people who drop things.
Roman Prokopyshyn had help from Claude fixing some problems with one of the VSPAERO example cases that hadn't been touched since before the thick/thin era.
Features:
- API code examples now serve as unit tests
- AI Audit for API completeness
- VSPAERO Outputs more stuff for dynamic analysis
Build System:
- Update AngelScript to v2.38
- Improve Python MANIFEST.in and packaging
Bugs:
- Fix AngelScript registration of vec3d and Matrix4d
- Fix thin shell inertia calculation
- Fix many small bugs identified by new unit tests
- Fix matrix spreadsheet UI referencing
- Fix DeltaFlatPlateDragArea for unsteady cases (thanks Jaime)
- Swept_Wing_API example VSPAERO script brought current (thanks Roman)
Reply all
Reply to author
Forward
0 new messages