bazel: Query with rbuildfiles frequently throws IllegalStateException

Description of the problem / feature request:

In our CI system, we run the following query to determine if any modified files are dependencies of any BUILD files:

 bazel query \
   --keep_going \
   --order_output=no \
   --universe_scope='//...' \
   "rbuildfiles($comma_sep_modified_files)"

Frequently, this results in an exception that looks like this:

java.lang.IllegalStateException: File:[/buildkite/builds/hello-world[source]]frontend/web/src/js/app/middleware/initialization.js=InMemoryNodeEntry{identity=728588638, value=null, lastChangedVersion=IntVersion: 388, lastEvaluatedVersion=IntVersion: 388, directDeps=null, reverseDeps=ReverseDeps{reverseDeps=[], singleReverseDep=false, dataToConsolidate=null}, dirtyBuildingState=FullDirtyBuildingState{dirtyState=CHECK_DEPENDENCIES, signaledDeps=-1, externalDeps=0, dirtyDirectDepIndex=0, lastBuildDirectDeps=GroupedList{elements=[FILE:[/buildkite/builds/hello-world]/[frontend/web/src/js/app/middleware/initialization.js]], size=1}, lastBuildValue=RegularFileArtifactValue{digest=00190f7f9e1cb320fe146036e972844242a6aa8353a2b9aff2740a536146f894, size=1459, proxy=ctime of 1559922057219 and nodeId of 1317526}}}
--
  | at com.google.common.base.Preconditions.checkState(Preconditions.java:507)
  | at com.google.devtools.build.skyframe.DelegatingWalkableGraph.getReverseDeps(DelegatingWalkableGraph.java:132)
  | at com.google.devtools.build.lib.query2.RBuildFilesVisitor.getVisitResult(RBuildFilesVisitor.java:80)
  | at com.google.devtools.build.lib.query2.ParallelVisitor$VisitTask.process(ParallelVisitor.java:217)
  | at com.google.devtools.build.lib.query2.ParallelVisitor$Task.run(ParallelVisitor.java:197)
  | at com.google.devtools.build.lib.concurrent.AbstractQueueVisitor$WrappedRunnable.run(AbstractQueueVisitor.java:387)
  | at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source)
  | at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source)

The affected filename differs on each build, but the stack trace is always the same, with the exception coming from RBuildFilesVisitor.

Feature requests: what underlying problem are you trying to solve with this feature?

Determining which targets may have been affected by a pull request.

Bugs: what’s the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

Unfortunately this is in a mid-sized, private monorepo which I can’t provide here. Running the bazel query provided above results in this exception around 10-20% of the time on our CI servers.

What operating system are you running Bazel on?

Linux

What’s the output of bazel info release?

# bazel info release
INFO: Invocation ID: 2537f98b-c39a-4e98-a4fd-90ff745c9416
release 0.26.1

Note however that I also saw this issue with Bazel 0.25.

What’s the output of git remote get-url origin ; git rev-parse master ; git rev-parse HEAD ?

1577a2caadd770afa708cf6e0eeea08cda579c34
bdfc59dde9a6eee44df21b66ba67ce027bd18e59

Have you found anything relevant by searching the web?

It is possible that #7294 is related but I’m not sure.