CI sdist fixes, port conda patch, fix ODR violation #2826
Conversation
fixes bug when cmake is installed to a path with a space in it Signed-off-by: Tom Jakubowski <[email protected]>
Also updates the cargo patch-job so it uses the .crate files produced earlier in the workflow instead of using git sources with metadata. This is closer to what happens when building and installing from a release sdist, which pulls .crate files from crates.io. Signed-off-by: Tom Jakubowski <[email protected]>
Change how the CMAKE_C_FLAGS_RELEASE and CMAKE_CXX_FLAGS_RELEASE variables are reset and used. They are now interpolated into CMAKE_CXX_FLAGS, placed before the EXTENDED_FLAGS so that EXTENDED_FLAGS and OPT_FLAGS can override the FLAGS_RELEASE. This ensures that RELEASE flags from the environment are used, while allowing them to be overridden by perspective's optimization flags. Also preserve values of `CMAKE_CXX_FLAGS` set by the builder. This is a port of a conda-feedstock patch by Isuru Fernando. <conda-forge/perspective-feedstock@9eab18c> Signed-off-by: Tom Jakubowski <[email protected]>
Signed-off-by: Tom Jakubowski <[email protected]>
this was an ODR violation: the same function was defined twice, once with `inline` specifier in utils.h, and once without `inline` in utils.cpp it happened to work on clang and gcc and msvc, most of the time, but broke in the conda build on Windows Signed-off-by: Tom Jakubowski <[email protected]>
|
test tag CI run, with a simulated 3.1.4 release commit: https://github.com/tomjakubowski/perspective/actions/runs/11677140432 |
|
Oneoff CI run which tests my WIP branch updating the perspective-feedstock repo to v3.1.4. The CI job uses perspective-client.crate and perspective-server.crate artifacts from the test tag build above. The build is using an sdist from a previous test tag run which I uploaded somewhere https://github.com/tomjakubowski/perspective-oneoff-tests/actions/runs/11678364614/job/32517697170 it wouldn't be too much trouble to update that workflow to instead download the sdist artifact from the same run it gets the .crates from. In short:
Then that workflow would be a decent way to sanity check future builds against conda (maybe with more platform coverage). Better still maybe to trigger (augmented versions of?) perspective-feedstock's workflows from our CI, supplying .crate and sdist artifacts from our workflows. |
Follow ups for #2811