Nighly feedstock build failed about conda-forge-nightly-controller

[jdblischak]

All the linux builds failing the linking test

[jdblischak]

Here's the error from the linux-64/openssl3/libcurl8 build:

$PREFIX/bin/make  -f CMakeFiles/link_test.dir/build.make CMakeFiles/link_test.dir/build
make[3]: Entering directory '$SRC_DIR/link_test_bld'
./ExampleExe
You are using TileDB version 2.21.0
make[3]: *** [CMakeFiles/link_test.dir/build.make:70: CMakeFiles/link_test] Segmentation fault (core dumped)
make[2]: *** [CMakeFiles/Makefile2:110: CMakeFiles/link_test.dir/all] Error 2
make[1]: *** [CMakeFiles/Makefile2:117: CMakeFiles/link_test.dir/rule] Error 2
make: *** [Makefile:137: link_test] Error 2
make[3]: Leaving directory '$SRC_DIR/link_test_bld'
make[2]: Leaving directory '$SRC_DIR/link_test_bld'
make[1]: Leaving directory '$SRC_DIR/link_test_bld'

WARNING: Tests failed for tiledb-2.21.0.2024_01_30-h4012625_0.conda - moving package to /home/conda/feedstock_root/build_artifacts/broken
TESTS FAILED: tiledb-2.21.0.2024_01_30-h4012625_0.conda

[jdblischak]

Ah, it's the same segmentation fault that is breaking geopandas (conda-forge/tiledb-feedstock#228 (comment)). tiledb is built against azure-core-cpp 1.10.3 but is co-installed with 1.11.0.

The good news is that this confirms it is a linux-only problem

[jdblischak]

The nightly builds are passing again. This is because they installed the new build of azure-core-cpp 1.10.3 that uses a max pin of x.x (conda-forge/azure-core-cpp-feedstock#12), and thus azure-core-cpp 1.10.3 is also installed at runtime for the test section

     azure-core-cpp:           1.10.3-h91d86a7_1           conda-forge