New CI pipline for fork with no changes (or any changes) fails in misc_jdk17 step

[nickbreen]

Setting up a new fork with Azure Pipelines CI fails consistently with this error in the misc_jdk17 step and test-misc.sh.

---

ERROR: Empty diff for 600920f...600920f.
Try pulling base branch (often main or master) into compare branch (often your feature branch).
Other causes of an empty diff include:

• The commit range is a no-op, such as a commit and its revert.
• HEAD is a merge commit and the merge took all its changes from its second argument.
• Your compare branch does not differ from your base branch.

---

https://dev.azure.com/nickbreen/checker-framework/_build/results?buildId=5&view=logs&j=03ef1a5a-d30a-57af-3a6d-62ddb34c1ad2&t=5d54b24c-372d-5954-02b2-e775bd9d239f&l=1814354

The diff is always the HEAD of the branch.

I expect that this message is the actual issue:

• Your compare branch does not differ from your base branch.

What is surprising/confusing is that this error occurred on the master branch of the fork... there is no compare branch.

It also occurred when creating a new branch, and when synchronising the fork from upstream.

[nickbreen]

As an aside, setting up a new "free" Azure account, one will need to explicitly request "parallel" jobs be enabled... initial run of the pipeline will fail with instructions.

[nickbreen]

I did find this in the docs: https://checkerframework.org/manual/developer-manual.html#azure-pipelines-bug-same-commit-hash

Azure Pipelines has a bug: whenever two CI jobs would run code with the same commit hash, it re-uses a previous execution result. This is a bug because the CI job's behavior may depend on the branch name and other factors that are independent of the commit hash. This means that you may see spurious successes or failures when your branch of (say) the jdk repository has the identical commit hash to some other branch that Azure Pipelines previous ran a job for.

But I'm not entirely sure if it is the same issue.

[nickbreen]

Just confirming that a commit avoids the error, as expected
.
A minor commit (updating a javadoc) still presents the same failure.

[nickbreen]

Actual failure:

https://dev.azure.com/nickbreen/checker-framework/_build/results?buildId=10&view=logs&j=03ef1a5a-d30a-57af-3a6d-62ddb34c1ad2&t=5d54b24c-372d-5954-02b2-e775bd9d239f&l=914026

Warning the raw log for that step is ~67MB

[nickbreen]

https://dev.azure.com/nickbreen/checker-framework/_build/results?buildId=16&view=logs&j=03ef1a5a-d30a-57af-3a6d-62ddb34c1ad2&t=5d54b24c-372d-5954-02b2-e775bd9d239f&l=12496

From https://github.com/nickbreen/checker-framework
* branch                master     -> FETCH_HEAD
git diff 600920f22049b5d55b8c1ade5fd23402dff016d0...5fede57c1ba8b28c4288ca043200b898f26555aa > /tmp/diff.txt
fatal: 600920f22049b5d55b8c1ade5fd23402dff016d0...5fede57c1ba8b28c4288ca043200b898f26555aa: no merge base
fatal: 600920f22049b5d55b8c1ade5fd23402dff016d0...5fede57c1ba8b28c4288ca043200b898f26555aa: no merge base
* (HEAD detached at origin/5fede57c1ba8b28c4288ca043200b898f26555aa)
  remotes/origin/5fede57c1ba8b28c4288ca043200b898f26555aa
  remotes/origin/master
fatal: 600920f22049b5d55b8c1ade5fd23402dff016d0...5fede57c1ba8b28c4288ca043200b898f26555aa: no merge base

[nickbreen]

Yet git diff 600920f22049b5d55b8c1ade5fd23402dff016d0...5fede57c1ba8b28c4288ca043200b898f26555aa is just fine.

[nickbreen]

Hunch: shallow clones of branches with more than one commit will not have a merge-base. Going to try a branch with a single squash commit.

I suspect that may have been the phantom green build I thought I saw in #6129 (comment)

https://dev.azure.com/nickbreen/checker-framework/_build/results?buildId=21&view=logs&j=03ef1a5a-d30a-57af-3a6d-62ddb34c1ad2&t=06a14f9f-12e3-5481-35b2-b69d36c91440&l=33 shows the --depth=1 argument.

[nickbreen]

Nope.

[nickbreen]

It seems that almost all the checkout steps have a fetchDepth set. The default apparently being 1.

Setting a conservative fetchDepth: 25 on the checkout steps that were missing the property seems to have avoided the issue.

[nickbreen]

Subject: [PATCH] everyone else has a fetchDepth so why can't i?
---
Index: azure-pipelines.yml
IDEA additional info:
Subsystem: com.intellij.openapi.diff.impl.patch.CharsetEP
<+>UTF-8
===================================================================
diff --git a/azure-pipelines.yml b/azure-pipelines.yml
--- a/azure-pipelines.yml	(revision 9bff96328aa6540c2aec94ce2747ecab3bfe2b8f)
+++ b/azure-pipelines.yml	(revision 717f705e0f18fab6c78b6ef35e3cfc767094c503)
@@ -160,6 +160,7 @@
   container: mdernst/cf-ubuntu-jdk11-plus:latest
   steps:
   - checkout: self
+    fetchDepth: 25
   - bash: ./checker/bin-devel/test-misc.sh
     displayName: test-misc.sh
 - job: misc_jdk17
@@ -168,6 +169,7 @@
   container: mdernst/cf-ubuntu-jdk17-plus:latest
   steps:
   - checkout: self
+    fetchDepth: 25
   - bash: ./checker/bin-devel/test-misc.sh
     displayName: test-misc.sh
 - job: misc_jdk20
@@ -179,6 +181,7 @@
   container: mdernst/cf-ubuntu-jdk20-plus:latest
   steps:
   - checkout: self
+    fetchDepth: 25
   - bash: ./checker/bin-devel/test-misc.sh
     displayName: test-misc.sh
 - job: typecheck_jdk11