@@ -143,14 +143,14 @@ unaffected build artifacts. That way, we'll only rebuild the minimum set of
143143changes.
144144
145145In real life, using a hermetic build system _ for tests_ does not actually
146- scale. The dependency graph of your repo becomes entwined with how much of your
147- repo gets tested. Change a deeply nested component often? Even though you only
148- want to deploy one out of all your services, you'll get full rebuilds of your
149- repo.
146+ scale for pre-merge checks . The dependency graph of your repo becomes entwined
147+ with how much of your repo gets tested. Change a deeply nested component often?
148+ Even though you only want to deploy one out of all your services, you'll get
149+ full rebuilds of your repo.
150150
151151I still think a hermetic build system is the way to go; but I wouldn't gate
152- merges on testing the full repo. Instead, let's give some flexibility to our
153- engineers.
152+ _ merges _ on testing the full repo. Instead, let's give some flexibility to our
153+ engineers.[ ^ post-merge ]
154154
155155If we have the service ` toy-srv ` , the developers of that service should be
156156comfortable in declaring what needs to pass in order for the service to be
@@ -293,3 +293,8 @@ left a lot of things out. I might write about this again in the future.
293293# ### Updates
294294* 2020-03-15: Improve phrasing as to why naively serializing merges does not
295295 scale ([commit](https://github.com/felixmulder/felixmulder.github.io/commit/6a484431e0b8acda13b4c01a20121c73df044280))
296+
297+ [^post-merge] : When the change has actually been merged, we could still run
298+ a more comprehensive suite of tests or even the full suite of tests and
299+ propose an automatic revert if there's been any breakage to dependencies.
300+ This is an awesome thing that a hermetic build system can provide!
0 commit comments