Silvio Traversaro
Silvio Traversaro
> @traversaro any updates from the Hunter's side of approach? If you fork works as expected, we could potentially utilize that as well, if upstream won't be too fast on...
Successful job use the `Version: 20241021.1.0` GitHub Actions windows-2022 image, while failing jobs are using Version: 20241113.3.0 and fail with: ~~~ 2024-11-18T09:43:13.1500390Z [31m│[0m D:\a\qpsolvers-eigen\qpsolvers-eigen>call :GetWin10SdkDir 2024-11-18T09:43:13.1500838Z [31m│[0m 2024-11-18T09:43:13.1501344Z [31m│[0m D:\a\qpsolvers-eigen\qpsolvers-eigen>call...
The non-determinism is due to https://github.com/actions/runner-images/issues/10978 .
The Windows failure is unrelated.
Last successful build: https://github.com/robotology/idyntree/actions/runs/15939515961 First failing build: https://github.com/robotology/idyntree/actions/runs/15950689204
The diff is quite clear: ~~~diff --- +++ @@ -11,7 +11,7 @@ ca-certificates 2025.6.15 h4c7d964_0 conda-forge cairo 1.18.4 h5782bbf_0 conda-forge cmake 4.0.3 hff78f93_0 conda-forge -cxx-compiler 1.9.0 h91493d7_0 conda-forge +cxx-compiler 1.10.0...
Also newer version fail, see https://github.com/robotology/idyntree/pull/1252 . As the crash is in the XML parsing code, see: ~~~ Stack Trace (from fault): [ 0] 0x00007ffb3fc43020 C:\Program Files\MATLAB\R2023b\bin\win64\MSVCP140.dll+00077856 Thrd_yield+00000272 [ 1]...
Installing such an old libxml2 is not easy: ~~~ error libmamba Could not solve for environment specs The following packages are incompatible ├─ glfw >=3.4 * is installable and it...
Indeed, installing the old libxml2 2.10.* solves the issue, see https://github.com/robotology/idyntree/pull/1252 . This indicates that the error is indeed related to libxml2. This is not easy to solve at all,...
By the way, proxqp developer provide this interesting euristics: https://simple-robotics.github.io/proxsuite/md_doc_22-ProxQP__api.html#OverviewWhichBackend . I would say that probably it could make sense to use the dense backend then, also by default? @LoreMoretti...