Uh oh!
There was an error while loading. Please reload this page.
GH-35658: [Packaging] Sync conda recipes with feedstocks - #35637
Conversation
Thanks for opening a pull request! If this is not a minor PR. Could you open an issue for this pull request on GitHub? https://github.com/apache/arrow/issues/new/choose Opening GitHub issues ahead of time contributes to the Openness of the Apache Arrow project. Then could you also rename the pull request title in the following format? or In the case of PARQUET issues on JIRA the title also supports: See also: |
h-vetinari
commented
May 17, 2023
@github-actions crossbow submit -g conda |
Revision: 8ef347b Submitted crossbow builds: ursacomputing/crossbow @ actions-ca516ae62e |
h-vetinari
commented
May 17, 2023
Interestingly, the failure of |
kou
commented
May 17, 2023
I don't know why we didn't need timezone database for Windows jobs before but we have a shell script to download timezone database: https://github.com/apache/arrow/blob/main/ci/scripts/download_tz_database.sh We're using the script on Windows too. |
h-vetinari
commented
May 18, 2023
Sounds like we should just add a dependency on the |
assignUser
commented
May 18, 2023
That sound's good. Thanks for taking care of the sync work, much appreciated! |
kou
commented
May 18, 2023
It may not work because it seems that |
h-vetinari
commented
May 18, 2023
Why does arrow depend on this in the first place? It's the first package I see that doesn't work with The tzdata stuff is an arrow 13 problem though. More urgently, in conda-forge/arrow-cpp-feedstock#1053, I currently have a failing test |
As a note, our nightly builds for wheels on Windows are also failing at the moment with the same error:
The first failure is from 10 days ago since this commit was introduced: |
kou
commented
May 19, 2023
Ah, sorry. I misread this problem. It seems that But tzdata path was hard-coded on Windows: https://github.com/apache/arrow/blob/main/cpp/src/arrow/vendored/datetime/tz.cpp#L265 So our code will not find tzdata installed by conda. Could you open a new issue for tzdata related problem? We can work on this as a separated issue. |
edit: strikethrough |
As I mentioned, the tzdata thing is much less urgent for us in conda-forge than the issue of that failing tests on the feedstock. I now opened a separate issue so this gets (hopefully...) more traction: #35728 |
jorisvandenbossche
commented
May 24, 2023
Arrow unfortunately can't work with the tz database from In other places we skip tests that requite tz database access on Windows, so that should be done for the new scalar test as well, I assume. |
jorisvandenbossche
commented
May 24, 2023
I opened #35735 to skip the test requiring a tz database on windows. |
I don't mind shipping arrow's vendored |
h-vetinari
commented
May 24, 2023
Absolutely, that's why it would be nice if this could be handled through a "proper" dependency like |
jorisvandenbossche
commented
May 24, 2023
That's probably technically possible to implement such transformation, but someone still needs to do the work for that. A potential easier path forward is to update date.h to work with the compiled database on Windows (HowardHinnant/date#564) |
h-vetinari
commented
Jul 24, 2023
@github-actions crossbow submit -g conda |
| - clangdev {{ llvm_version }} | ||
| - llvmdev {{ llvm_version }} | ||
| - cython | ||
| - cython <3 |
There was a problem hiding this comment.
I wanted to wait for #36745 (and backport it 12.0.1), but that PR has not landed as quickly as I hoped.
Revision: e9e6605 Submitted crossbow builds: ursacomputing/crossbow @ actions-b1a7f0eaed |
h-vetinari
commented
Jul 24, 2023
@github-actions crossbow submit -g conda |
Revision: 17fc588 Submitted crossbow builds: ursacomputing/crossbow @ actions-dd4f8afbf0 |
@kou@pitrou@raulcd@assignUser@jorisvandenbossche |
assignUser
left a comment
There was a problem hiding this comment.
Great work, thanks @h-vetinari ! I will merge in a day or so if there are not objections
| EXTRA_CMAKE_ARGS="${EXTRA_CMAKE_ARGS} -DLLVM_LINK_EXECUTABLE=${BUILD_PREFIX}/bin/llvm-link" | ||
| sed -ie "s;protoc-gen-grpc.*$;protoc-gen-grpc=${BUILD_PREFIX}/bin/grpc_cpp_plugin\";g" ../src/arrow/flight/CMakeLists.txt | ||
| sed -ie 's;"--with-jemalloc-prefix\=je_arrow_";"--with-jemalloc-prefix\=je_arrow_" "--with-lg-page\=14";g' ../cmake_modules/ThirdpartyToolchain.cmake | ||
| sed -ie 's;"--with-jemalloc-prefix\=je_arrow_";"--with-jemalloc-prefix\=je_arrow_" "--with-lg-page\=16";g' ../cmake_modules/ThirdpartyToolchain.cmake |
There was a problem hiding this comment.
We can use -DARROW_JEMALLOC_LG_PAGE=16 instead of rewriting ThirdpartyToolchain.cmake.
There was a problem hiding this comment.
That sounds great, I'll change that on the feedstock and in the next sync
After merging your PR, Conbench analyzed the 6 benchmarking runs that have been run so far on merge-commit 8503c86. There were no benchmark performance regressions. 🎉 The full Conbench report has more details. |
…e#35637) Corresponds to conda-forge/arrow-cpp-feedstock#1053 for pyarrow (one failing test on our infra that needs debugging), as well as the state of the feedstock for r-arrow after conda-forge/r-arrow-feedstock#65 * Closes: apache#35658 Authored-by: H. Vetinari <h.vetinari@gmx.com> Signed-off-by: Sutou Kouhei <kou@clear-code.com>
Corresponds to conda-forge/arrow-cpp-feedstock#1053 for pyarrow (one failing test on our infra that needs debugging), as well as the state of the feedstock for r-arrow after conda-forge/r-arrow-feedstock#65