Instead of a local database, this script now imports the data
into the shared postgres database. this way the data can be used
for other queries as well.
feel free to extend the data model if you
need additional columns.
This is a wild attempt to fix the pre-commit CI which has been failing
for a few hours. I think the authors of 1f2c851 are not available right
now, so I'm taking the freedom to try this out. I hope I'm not overstepping
any boundary.
using metadata tags set in #298 finds and cancels existing builds before
starting a new one.
One caveat is that no result is reported back to Phabricator for the
cancelled build. That should not be an issue in the normal usecase.
For #278
Those bindings requires extra dependencies (like Python, pybind11)
and clang-tidy fails to analyze the source files if they are not
configured properly.
Now "report" step combines result in a uniform way and processes unit test
results XML output. It works for sub-builds only started from the 'premerge'
pipeline, i.e. non-recursive. One downside is that now one has to wait until
all jobs have finished.
- Add instructions to setup python environment
- added option to do full report cycle but not call Phabricator
- use "annotations" to show build status. That lifts the need to filter ninja
and other output (thus `ph_no_filter_output` param removed) and output
everything. That is nice as script failures no longer lead to loss of logs.
- improved annotate() usability
- misc fixes
- improve messaging and add instructions how to fix the patch
- fix application of patch stack. Now base commit and order of patches
should be correct
- set origin to llvm-project fork to avoid accidental operations on origin
- use annotations to make issues visible
- create commits with original author of change
- add patches we tried to apply to artifacts
- patching / commits works locally if --push-branch=false is set
- misc renames / wordings
fixes#192
additionally:
- install python libs after scripts checkout, so we don't need to
rebuild images and restart agents only to add a new python dependency
- updated lib versions
- similar scripts checkout in steps
Previous approach had drawbacks:
- every step had to implement exporting of results in fixed format
- if step failed then failure will not be detected
Now report step will fetch results directly from Buildkite.
Agents have to be updated to have BUILDKITE_API_TOKEN env.
Previously list of projects was resolved when job run at target OS.
Now project resolution is moved to the premerge pipeline and logic
should be there.
Now we run a pre-commit hook that tries to kill any process that locked
a file in target directory.
Updated timeout interval on windows from 120 to 90 minutes.
Fixes#243
Now it's possible to allow sub-projects to define own checks and skip
"generic" ones.
To properly accomodate affected projects that might not have special
treatment we:
1. extend the set of affected projecs with dependent (e.g. add 'libc' if
'clang' was modified)
2. add custom steps for projects that define own workflow. At the moment
it's only libcxx and it has a custom trigger pipeline so it's noop.
3. add dependent projects and run generic check on them.
To illustrate: imagine that we have a dependency graph:
llvm -> clang -> openmp
and only clang was modified in a diff; also clang defines own checks.
Thus list of affected projects will be [clang, openmp].
After adding custom checks and removing their projecst: [openmp].
After adding dependencies: [llvm, clang, openmp]. Generic linux /
windows checks will be run on thouse 3 projects.
So as you can see in some scenarios projects with custom checks will
still go through generic checks.
Note that clang-format and clang-tidy checks are run only for "generic"
checks at the moment.
That gives us a way to test e.g. pull requests by specifying
ph_scripts_refspec="pull/123/head"
Also keep "ph_" prefix to remove confusion we had between
"scripts_branch" and "ph_scripts_branch" before.
Update docs and moved "testing" sections up.
some compiler-rt files contain architecture-specific assembly code or
complex macroses which clang-tidy can't process properly.
In case of assembly code: we can get "unknown register name 'a7' in asm"
In sanitizer_common_syscalls.inc we have issuels like:
"clang-tidy: error: unknown type name '__sanitizer_iovec'"
- moved linux agents to a new node pool (basically renamed old one)
- removed some out of date scripts or moved them to playlists
- removed shell scrips invoked from jenkins
- minor docs updates
Some python classes methods are not used anymore and will be cleaned up
separately.
https://ccache.dev/manual/3.7.11.html#_configuration uses CCACHE_DIR to set
cache directory. CCACHE_PATH has a different meaning.
+ removed LLVM_CCACHE_DIR ccache argument to use CCACHE_DIR from env.
+ print additional info about ccache config and machine readable stats.
Both files include GPU header files for which clang implicitly adds include directories. However, clang-tidy does not and therefore we get compile errors (i.e. not diagnostics that we could disable with NOLINT).
- Fix for #207. Previously grep was exiting with 1 and fail build if all
lines from ninja are filtered. Also piping might hide exit code of the
initial command. Now "tee" and "grep" is implemented on python level
and we get exit code of the original command.
- Fixed default value of "scripts_branch", print scripts commit;
- Added option to filter ninja output (on by default as before).
- Copied cache usege reporting and error handling from "master" branch
build to diff checks (useful for debugging).
apt.llvm.org signature has changed, we also should use debian:stable
(=buster) as a base image to correctly install packages.
Removed jdk from the base image as it was used only by Jenkins.
Log packages version for every build.
Related issues: #112
libclc implements OpenCL C, which is not actual C/C++, and implements an
external API with different naming conventions.
As clang-tidy cannot parse the source[0], just blacklist all of libclc's
sources for now.
[0]: https://buildkite.com/llvm-project/premerge-checks/builds/695
- configured sccache on buildkite windows machines
- final result is sent to phabricator from "summary.py" that waits for both builds to complete
- extracted "add_url_artifact" to a runnable script
- reorganized code and fixed some of TODOs
I have tried to install new windows machine for buildkite and updated
some scrips and docs on along the way.
- Updated base image for k8s agent installation as it gave a warning that
previous version was old.
- Now buildkite secret is configured in a powershell script along with
tags (and possible other parameters).
- Split "windows_agent_start.ps1" to "..jenkins" and "..buildkite" as
some parameters are different.
- Created a "windows-development" machine in GCP stat can be resumed
to build docker images / expriment.
Harbormaster can send a request in url encoded form while buildkite
expects a POST request with json. Instead of modifying harbormastar or
buildkite this adds a simple proxy that accepts url encoded form and
creates a request that buildkite expects. To avoid potential abuse,
nginx asks for simple http auth credentials stored in harbormaster.
All build parameters passed by proxy are put into build metadata and as
ph_* env variables available during the build.
Secrets involved:
- harbormastert knows http-auth to proxy (stored in privatly and in k8
buildkite/http-auth as auth file);
- proxy knows buildkite api token (mine atm);
- build agent knows conduit API token (mine atm),
and SSH key of llvm-premerge-tests-bot (in k8 buildkite/github-ssh).
Sample build: https://reviews.llvm.org/harbormaster/build/64828/8/https://buildkite.com/llvm-project/premerge/builds/48
