vtk 9.0.0.rc1 test often fail because of limited connections
When trying to build vtk on conda-forge using CI, I am running into frequent failures like the following:
[1351/13631] Generating D:/bld/vtk_1584607138539/work/build/ExternalData/Filters/HyperTree/Testing/Data/Baseline/TestHyperTreeGridBinary2DAxisClipPlanes.png
FAILED: ExternalData/Filters/HyperTree/Testing/Data/Baseline/TestHyperTreeGridBinary2DAxisClipPlanes.png-hash-stamp ExternalData/Filters/HyperTree/Testing/Data/Baseline/TestHyperTreeGridBinary2DAxisClipPlanes.png
cmd.exe /C "cd /D %SRC_DIR%\build && %BUILD_PREFIX%\Library\bin\cmake.exe -Drelative_top=D:/bld/vtk_1584607138539/work/build -Dfile=D:/bld/vtk_1584607138539/work/build/ExternalData/Filters/HyperTree/Testing/Data/Baseline/TestHyperTreeGridBinary2DAxisClipPlanes.png -Dname=D:/bld/vtk_1584607138539/work/Filters/HyperTree/Testing/Data/Baseline/TestHyperTreeGridBinary2DAxisClipPlanes.png -Dexts=.sha512 -DExternalData_ACTION=fetch -DExternalData_CONFIG=D:/bld/vtk_1584607138539/work/build/VTKData_config.cmake -P D:/bld/vtk_1584607138539/_build_env/Library/share/cmake-3.16/Modules/ExternalData.cmake"
-- Fetching "https://data.kitware.com/api/v1/file/hashsum/SHA512/3cc237eb171341b339c859174bf00d4efb4047f6fc89efd60c5c87720323f15e63458a8be01d210c29fd150cee4e3a5c0a3df7ad26e3a290660c4a8607a5b042/download"
-- Fetching "https://www.vtk.org/files/ExternalData/SHA512/3cc237eb171341b339c859174bf00d4efb4047f6fc89efd60c5c87720323f15e63458a8be01d210c29fd150cee4e3a5c0a3df7ad26e3a290660c4a8607a5b042"
CMake Error at D:/bld/vtk_1584607138539/_build_env/Library/share/cmake-3.16/Modules/ExternalData.cmake:1121 (message):
Object
SHA512=3cc237eb171341b339c859174bf00d4efb4047f6fc89efd60c5c87720323f15e63458a8be01d210c29fd150cee4e3a5c0a3df7ad26e3a290660c4a8607a5b042
not found at:
https://data.kitware.com/api/v1/file/hashsum/SHA512/3cc237eb171341b339c859174bf00d4efb4047f6fc89efd60c5c87720323f15e63458a8be01d210c29fd150cee4e3a5c0a3df7ad26e3a290660c4a8607a5b042/download ("HTTP response code said error")
https://www.vtk.org/files/ExternalData/SHA512/3cc237eb171341b339c859174bf00d4efb4047f6fc89efd60c5c87720323f15e63458a8be01d210c29fd150cee4e3a5c0a3df7ad26e3a290660c4a8607a5b042 ("Couldn't connect to server")
So far, these are always under Windows, but this may simply be because of the order of the CI tests or how long it takes each to set up before these downloads start. I saw similar issues when I tried to git clone
the repo and had to switch to pulling data from GitHub instead of gitlab.kitware.com.
Any suggestions on how to make this more robust? It's not good for us when CI fails and has to be restarted because of server condition issues, particularly on branch merges, where no one may be watching to make sure the operation completed successfully.