Repository navigation
Add Arrow and Parquet C++ to the Rocky build images - #2155
Merged
MisterRaindrop merged 1 commit intoOct 21, 2026
Merged
Conversation
contrib/datalake_fdw builds against Arrow and Parquet C++. Install them in the Rocky 8, 9 and 10 build images from the Arrow project's own repository, pinned so the version moves only with a change here: 22.0.0 on Rocky 8, the last release whose headers are C++17 (from 23 on they need C++20, which gcc 8 does not have), and 26.0.0 on Rocky 9 and 10. The repository is left disabled afterwards, as EPEL is. The release package goes in with every repository disabled: it needs nothing from them, and loading their metadata under QEMU took eight minutes of an arm64 build. Testinfra checks that pkg-config reports the pinned version of both. The arm64 half of each build runs under QEMU, and the Rocky builds already took 35 to 60 minutes before Arrow joined them, so the job timeout goes from 60 to 90 minutes. The Ubuntu images are left alone: Apache's libarrow-dev pulls in libcurl4-openssl-dev, which replaces the libcurl4-gnutls-dev the deb packaging depends on.
tuhaihe
approved these changes
Oct 20, 2026
yjhjstz
approved these changes
Oct 20, 2026
Alena0704
approved these changes
Oct 20, 2026
5 of 13 tasks
yjhjstz
pushed a commit
that referenced
this pull request
Oct 22, 2026
The Rocky build images now ship Arrow and Parquet C++ themselves (#2155): 22.0.0 on Rocky 8 and 26.0.0 on Rocky 9 and 10, with the Arrow repository left disabled. The ic-datalake-fdw step still installed its own copy, so once the images were rebuilt it failed before building anything: on Rocky 9 and 10 EPEL's libarrow-devel conflicts with the image's arrow-devel over the same headers, and on Rocky 8 the pinned 17.0.0 is no longer reachable. The step now installs nothing and builds against the image's Arrow. Arrow 23 moved its headers to C++20 (std::span, std::bit_width), so the Makefile picks -std=c++20 when Arrow is 23 or newer and keeps C++17 below that. C++20 deprecates ++ and -- on a volatile, which the server's own HOLD_INTERRUPTS does, so -Wvolatile is off with it. Arrow 26 removed the FileReader::GetRecordBatchReader overload that returns its reader through an out parameter. The Result-returning form exists from Arrow 19, so the reader uses it from there and keeps the old spelling for older versions.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Installs Arrow and Parquet C++ in the Rocky 8, 9 and 10 build images, so
contrib/datalake_fdw(#2044) can build against them without each CI job installing them itself.pkg-config --modversionreports the pinned version forarrowandparquet.The Ubuntu images are not changed: Apache's
libarrow-devpulls inlibcurl4-openssl-dev, which removes thelibcurl4-gnutls-devthat the deb packaging (devops/build/packaging/deb/ubuntu*/control) depends on. That needs its own discussion.Type of Change
Test Plan
Why the timeout changes. The first fork run had Rocky 9 cancelled at 60 minutes while exporting its arm64 image. Its Arrow layer took 12 minutes under QEMU, eight of them spent loading repository metadata for a release package that needs none; it now installs with
--disablerepo='*'. But the part of the build that does not involve Arrow varies a lot on its own: the arm64 main layer took 30 to 41 minutes across these runs, and the Rocky jobs took 35 to 60 minutes before this change. So 60 leaves no margin.Impact
Dependencies:
Image size grows by about 440–540 MB per Rocky image. Most of it is the static libraries in
arrow-devel(libarrow_bundled_dependencies.a150 MB,libarrow.a70 MB) andboost-devel, whichthrift-develrequires.datalake_fdwlinks the shared libraries, so the static ones could be deleted in the same layer if the size matters; I left them so the image stays a plain package install.Checklist
Additional Context
Once the published images carry Arrow, a follow-up changes the datalake job in
build-cloudberry.ymlto use it and drops the Arrow install there.