ianoc / bazelfe

Set of tools to handle errors/work with bazel invocations

Bazel runner

This is archived

-- Last checked this still worked fine for purpose, but original primary user is gradually moving off. So doesn't have a solid user. Archiving to indicate its no longer being maintained and may not work with newer versions of bazel.

Goal/motivation

Be a suite of tools to provide ultimately a different frontend around bazel. Initially this is something to be injected to wrap calls for build/test operations in bazel to repair common issues.

Requirements

• Ability to run/provide a CI job or some means to run the index building cod.
• Not be consuming the BEP from bazel directly. The tool hooks in and tells bazel to send the BEP out to it to sniff on the results.

Using it

The best way to see it in action is to visit the examples folder and have a look at what it does.

1. Configure a CI job to run the indexer, it should produce a binary output
2. Store the output in a location which is fetchable by your developers/users
3. Expose buildozer at a path(fetch or via https://doc.rust-lang.org/std/macro.include_bytes.html or https://github.com/pyros2097/rust-embed we could probably embed in the release to ease distribution.)
4. From the examples you need to install: -> Some code/bash script (could be built into the launcher in future?) to fetch the index to provide -> Bash script for tools/bazel to alloow hooking into the bazel commands and delegating to the bazel-runner application
5. Run it

Other things: We also include/have a small script to measure how well it can do for you/potentially handle targets with unused dependencies. the slow_unused_deps.sh script will remove all dependencies from a target then try build it again. If the above is all working right, hopefully like magic it should just recover + build ok.

TODO:

• Bazel runner that can wrap bazel
• x ] JVM Indexer to find/index all jvm producing targets
  • Investigate TUI/tabbed interface when running tooling so that the bazel stdout/stderr can be preserved/viewed but not swamp things like indexing
  • Investigate using an aspect to gather the index information --- skipping this one for now, since current system seems to work well.
• Example project
• All scripts in the right place
• Integration for auto formatting handling for java/scala
• Investigate persistant daemon mode:
  • When file changes rebuild the target that owns it
  • When the above is successful run tests that directly ddepend on the rebuilt target
  • Optionally run all tests that transitively depend on the target
• Build UI experiments using the TUI library to show better histograms/data while building.
• Web interface?

Bazel runner

basic

USAGE:
    bazel-runner [OPTIONS] <passthrough-args>... --buildozer-path <buildozer-path>

ARGS:
    <passthrough-args>...

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
        --bind-address <bind-address>
            [env: BIND_ADDRESS=]

        --buildozer-path <buildozer-path>
            [env: BUILDOZER_PATH=]

        --config <config>
        --disable-action-stories-on-success <disable-action-stories-on-success>
            [env: DISABLE_ACTION_STORIES_ON_SUCCESS=]

        --index-input-location <index-input-location>
            [env: INDEX_INPUT_LOCATION=]

Options:

• Bind adddress, optional, change the ip/port we bind our GRPC server to that we tell bazel to use for BEP
• buildozer path required for buildozer operations to allow making changes/sniffing dependencies in build files
• passthrough args, what the user called bazel with, e.g.: /path/to/bazel-real build --flag --flag src/main/blah:wer