- Feb 27, 2025
-
-
Nao Hashizume authored
-
Stan Hu authored
Debian bookworm ships with glibc 2.36, but many of our tests use Ubuntu 22.04, which uses glibc 2.35. Use the Omnibus Ubuntu 22.04 build image, which contains all the right Ruby and baseline Go versios. Relates to gitlab-org/quality/engineering-productivity/master-broken-incidents#10724
-
Kev Kloss authored
This fixes a bug in the gitlab-runner setup task, which used the wrong variable, so it will update the `gdk` directory instead of the `gitlab-runner`.
-
Peter Leitzen authored
-
- Feb 26, 2025
-
-
- Feb 25, 2025
-
-
-
Omar Qunsul authored
Changelog: other Addressing gitlab#513075
-
Kev Kloss authored
This improves the behavior when connecting to gdk.test via HTTP when you have HTTPS enabled. Instead of responding with a 497 status code and error page, we redirect to the same URI but using HTTPS.
-
-
- Feb 24, 2025
-
-
Patrick Steinhardt authored
In 6d60771a (Merge branch 'pks-build-git-via-meson' into 'main', 2025-02-10), we have started to migrate the Gitaly build to start using Meson for building Git. One oversight though is that CI uses a different set of dependencies when compiling Gitaly, and this set of dependencies was not adjusted, consequently breaking the CI build. Fix this by installing necessary prerequisites in our CI compile jobs.
-
Sylvester Chin authored
-
Doug Barrett authored
-
- Feb 21, 2025
- Feb 19, 2025
-
-
Evan Read authored
-
- Feb 18, 2025
-
-
Peter Leitzen authored
List of files and SHAs when the last use was removed. - verify-one-line-install: 30193441 - remove-empty-file: b6b1de24 - ask-to-restart: 434bbb7e Makefile target `clean-gdk-root` was added in 2023 to remove a stale file. This is no longer necessary in 2025.
-
Kev Kloss authored
-
Peter Leitzen authored
Running `rake gdk.example.yml` became flaky in lefthook after we've disabled nginx's sendfile for Mac. This commit ensures that the example config is always generated as Linux x86 to side-step these issues.
-
Peter Leitzen authored
-
- Feb 17, 2025
-
-
-
Peter Leitzen authored
We've seen "400 Bad request" responses on MacOS with sendfile enabled. See #2413
-
- Feb 14, 2025
-
-
- Feb 12, 2025
-
-
-
Sangwoo Han authored
-
-
James Liu authored
Several flags were removed in v1.63.0 which don't appear to have any affect on the functionality. These removals aren't even called out as breaking changes in the release changelog [1]. Remove the flags from the Procfile. [1] https://github.com/jaegertracing/jaeger/releases/
-
James Liu authored
$MACHTYPE is currently being used to determine the host machine's CPU architecture. We translate from this value to a standard form to construct the download URL for Jaeger. Unfortunately, $MACHTYPE is not guaranteed to reflect the actual architecture of the machine. It's more likely to return the architecture of the machine that compiled the shell that it's executing in [1]. Instead, invoke `uname -m` to determine the architecture. The output is defined by POSIX and will work across systems reliably. Add an entry for `arm64`, as this is what macOS on Apple Silicon reports. [1] https://apple.stackexchange.com/questions/467846/machtype-environment-variable-on-arm
-
James Liu authored
Checksums for the Jaeger artifacts haven't been updated for some time. This causes checksum validation errors when newer versions of Jaeger are specified in the `gdk.yml` configuration file. Add the latest 1.x release [1] of Jaeger to the list of checksums. We don't want to upgrade to 2.x yet. [1] https://github.com/jaegertracing/jaeger/releases
-
James Liu authored
The query.host-port flag was deprecated and replaced [1] with the query.http-server.host-port flag: > (deprecated, will be removed after 2020-12-31 or in release v1.21.0, > whichever is later) see --query.http-server.host-port and > --query.grpc-server.host-port Use the new flag instead, because recent versions of the Jaeger CLI will error if given the old flag. [1] https://www.jaegertracing.io/docs/1.21/cli/ # Conflicts: # support/templates/Procfile.erb
-
- Feb 11, 2025
-
-
Dylan Griffith authored
In gitlab!176742 we are adding a `duo_worklfow.debug` setting to enable debug logging for the executor. This allows us to control that setting via the GDK and defaults it to `true`, because most people working in the GDK will want the verbose debug logging for the executor.
-
- Feb 07, 2025
-
-
Nao Hashizume authored
Running `bundle exec rake mise:migrate` directly doesn't load environment variables from .gdkrc, so we need a make target since Makefile sets SHELL = bin/gdk-shell.
-
Currently, when you enable Prometheus it cannot scrape the GDK as it can't find a route to the GDK. This change enables this and also clarifies the setup docs.
-
- Feb 06, 2025
-
-
-
Stan Hu authored
The Developer Experience team is trying to improve the mapping of what code actually runs when end-to-end tests execute. They saw a lot of noise in the coverband data showing cronjobs being mapped to unrelated tests. Being able to run groups of jobs in a different process should help improve these mappings. Now that we have gitlab!180136, we can route all cronjobs to their own queue. Relates to gitlab-org/quality/quality-engineering/team-tasks#2993
-
Patrick Steinhardt authored
In the preceding commit we have started to install required dependencies to build Git with Meson. Set the `USE_MESON` build flag to ask Gitaly to use the new build infrastructure.
-
Patrick Steinhardt authored
We have introduced new build infrastructure into Git using the Meson build system. The intent is that Meson will replace autoconf-based and CMake-based builds, and eventually also the Makefile. We want to start dogfooding this build in Gitaly to detect issues and lacking features with the new infrastructure. Install both Meson and Ninja so that Gitaly can do so. Fixes #2354
-
- Feb 05, 2025
-
-
Peter Leitzen authored
Usually, `gdk update` does not stop or start services. This is done by `gdk restart` later.
-