fix: Rebuild cached failures on proper workers

Author: Erethon

buildbot_nix/buildbot_nix/__init__.py

@@ -348,16 +348,20 @@ async def run(self) -> int:
                 msg.append(f"  failed build: {url}")
                 error_log.addStderr("\n".join(msg) + "\n")
             return util.FAILURE
-        self.build.addStepsAfterCurrentStep(
-            nix_build_steps(
-                self.project,
-                self.worker_names,
-                self.post_build_steps,
-                self.branch_config_dict,
-                self.outputs_path,
-                self.show_trace,
+        self.build.addStepsAfterCurrentStep([
+            Trigger(
+                name="Rebuild cached failure",
+                waitForFinish=True,
+                schedulerNames=[f"{self.project.project_id}-rebuild"],
+                haltOnFailure=True,
+                flunkOnFailure=True,
+                sourceStamps=[],
+                alwaysUseLatest=False,
+                updateSourceStamp=False,
+                copy_properties=["attr", "system", "branch", "revision"],
+                set_properties={"reason": "rebuild"},
             )
-        )
+        ])
         return util.SUCCESS
 
 
@@ -1075,6 +1079,11 @@ def config_for_project(
                 name=f"{project.project_id}-nix-register-gcroot",
                 builderNames=[f"{project.name}/nix-register-gcroot"],
             ),
+            # this is triggered from cached failure rebuilds
+            schedulers.Triggerable(
+                name=f"{project.project_id}-rebuild",
+                builderNames=[f"{project.name}/nix-eval"],
+            ),
             # allow to manually trigger a nix-build
             schedulers.ForceScheduler(
                 name=f"{project.project_id}-force",