Extending the Build Pipeline
Builder uses the concept of a “Build Pipeline” to build a project. The build pipeline consistes of multiple “phases” and build “stages” run in a given phase.
For example, in the Ide.PipelinePhase.DOWNLOADS
phase, you might have a stage
that downloads and installs the dependencies for your project. The Flatpak
extension does this when building Flatpak-based project configurations.
The various phases of the build pipeline are the following and are always executed in exactly this sequence up to the requested phase.
Build Phases
Ide.PipelinePhase.PREPARE
is the first phase of the build pipeline. Use this to create necessary directories and other preparation steps.
Ide.PipelinePhase.DOWNLOADS
should be used to download and cache any build artifacts that are needed during the build.
Ide.PipelinePhase.DEPENDENCIES
should build any dependencies that are needed to successfully build the project.
Ide.PipelinePhase.AUTOGEN
should generate any necessary project files. Contrast this with theIde.PipelinePhase.CONFIGURE
phase which runs the configuration scripts.
Ide.PipelinePhase.CONFIGURE
should run configuration scripts such as./configure
,meson
, orcmake
.
Ide.PipelinePhase.BUILD
should perform the incremental build process such asmake
orninja
.
Ide.PipelinePhase.INSTALL
should install the project to the configured prefix.
Ide.PipelinePhase.EXPORT
should be used to attach export hooks such as building a Flatpak bundle, Debian, or RPM package.
Additionally, there are phases which have special meaning.
Ide.PipelinePhase.BEFORE
can be XOR’d with any previous phase to indicate it should run as part of the phase, but before the phase has started.
Ide.PipelinePhase.AFTER
can be XOR’d with any previous phase to indicate it should run as part of the phase, but after the phase has completed.
Ide.PipelinePhase.FINISHED
indicates that a previous build request has finished.
Ide.PipelinePhase.FAILED
indicates that a previous build request has failed.
Creating a Build Stage
To add a build stage, we start by creating a build pipeline addin. When it loads we will register our stage in the appropriate phase.
# my_plugin.py
import gi
from gi.repository import GObject
from gi.repository import Ide
class MyPipelineStage(Ide.Object, Ide.PipelineStage):
def do_execute(self, pipeline, cancellable):
"""
This is a synchronous build stage, which will block the
main loop. If what you need to do is long running, you
might consider using do_execute_async() and
do_execute_finish().
"""
print("Running my build stage!")
def do_clean_async(self, pipeline, cancellable, callback, data):
"""
When the user requests that the build pipeline run the
clean operation (often before a "rebuild"), this function
will be executed. Use it to delete stale directories, etc.
"""
task = Gio.Task.new(self, cancellable, callback)
task.return_boolean(True)
def do_clean_finish(self, task):
return task.propagate_boolean()
def do_query(self, pipeline, cancellable):
"""
If you need to check if this stage still needs to
be run, use the query signal to check an external
resource.
By default, stages are marked completed after they
run. That means a second attempt to run the stage
will be skipped unless set_completed() is set to False.
If you need to do something asynchronous, call
self.pause() to pause the stage until the async
operation has completed, and then call unpause()
to resume execution of the stage.
"""
# This will run on every request to run the phase
self.set_completed(False)
def do_chain(self, next):
"""
Sometimes, you have build stages that are next to
each other in the pipeline and they can be coalesced
into a single operation.
One such example is "make" followed by "make install".
You can detect that here and reduce how much work is
done by the build pipeline.
"""
return False
class MyPipelineAddin(GObject.Object, Ide.PipelineAddin):
def do_load(self, pipeline):
stage = MyPipelineStage()
phase = Ide.PipelinePhase.BUILD | Ide.PipelinePhase.AFTER
stage_id = pipeline.attach(phase, 100, stage)
# track() can be used to auto-unregister the phase when
# the pipeline is removed.
self.track(stage_id)