[MAINT] Add conda-recipe for conda-forge only dpctl builds - #2367
[MAINT] Add conda-recipe for conda-forge only dpctl builds#2367ndgrigorian wants to merge 1 commit into
Conversation
aligned with dpctl-feedstock on conda-forge's recipe and scripts
|
coverage: 74.462%. remained the same — add-conda-forge-only-ci-tests into master |
ndgrigorian
changed the title
| # let the public CI build without IPO, which needs more resources than the | ||
| # GitHub Actions runners provide | ||
| script_env: | ||
| - OVERRIDE_INTEL_IPO # [win] |
There was a problem hiding this comment.
We are not using that in the feedstock. And I believe we probably no need that for dpctl build anymore, since it's lightweighted now.
| --no-index ^ | ||
| --prefix %PREFIX% ^ | ||
| -vv | ||
| if %ERRORLEVEL% neq 0 exit 1 |
There was a problem hiding this comment.
Here the check is effectively dead — it always sees the pre-loop errorlevel. To actually catch a failing pip install inside the block we need:
| if %ERRORLEVEL% neq 0 exit 1 | |
| if errorlevel 1 exit 1 |
Probably needs to be backported to the feedstock also.
There was a problem hiding this comment.
While at top level, both work. if %ERRORLEVEL% neq 0 is actually more precise than if errorlevel 1 because the latter is a >= test and also matches unusual negative codes differently. This is why the run_test.bat scripts (all top-level statements) are fine as written.
| cxx_compiler: # [linux] | ||
| - gxx # [linux] | ||
| cxx_compiler_version: # [linux] | ||
| - '14' # [linux] |
There was a problem hiding this comment.
to match the feedstock
| - '14' # [linux] | |
| - '15' # [linux] |
|
View rendered docs @ https://intelpython.github.io/dpctl/pulls/2367/index.html |
3 participants
This PR proposes adding a workflow for conda-forge builds of the project, which test that the Conda packages for the compiler dependencies (and NumPy) brought in purely from conda-forge work with dpctl