Improve determining the lines added by a branch

Since the list of lines added to Git-tracked text files in a given
branch is not part of the Danger DSL [1], it is determined using custom
code in dangerfile.py.  The current implementation of that logic is less
than perfect as it examines the diff between the current tip of the
target branch and the source branch rather than the diff between the
merge base of the two branches and the source branch.  Consider a Git
history like this:

    *   F (target)
    ...
    *   E
    *   D
    *   C
    | * B (source)
    |/
    *   A (merge base)

If danger-python or Hazard are run for commit B, the current logic for
determining the list of added lines in dangerfile.py examines the diff
between commits F and B rather than between commits A and B.  Therefore,
the added_lines() function returns not just the lines added by commit B
on top of commit A, but also the list of lines that were removed between
commits A and F, which leads to confusing results.

Fix by using the triple-dot diff operator in the Git invocation whose
output is used as the source of information for determining the list of
lines added by a given branch.

Since Hazard fetches the target branch itself when it is run, remove the
explicit "git fetch" invocation that fetches the target branch from
GitLab (shortening its local history to a single commit in the process)
before "git diff" is invoked.

[1] https://danger.systems/js/reference.html#GitDSL

diff --git a/dangerfile.py b/dangerfile.py
index ee77f13cd1156e9e822c03f382e7def98f1ea0ef..531e05bb185ddf1093334a9f90b173ee4222c309 100644 (file)
--- a/dangerfile.py
+++ b/dangerfile.py
@@ -23,11 +23,13 @@ import gitlab
 def added_lines(target_branch, paths):
     import subprocess
 
-    subprocess.check_output(
-        ["/usr/bin/git", "fetch", "--depth", "1", "origin", target_branch]
-    )
+    # Hazard fetches the target branch itself, so there is no need to fetch it
+    # explicitly using `git fetch --depth 1000 origin <target_branch>`.  The
+    # refs/remotes/origin/<target_branch> ref is also expected to be readily
+    # usable by the time this file is executed.
+
     diff = subprocess.check_output(
-        ["/usr/bin/git", "diff", "FETCH_HEAD..", "--"] + paths
+        ["/usr/bin/git", "diff", f"origin/{target_branch}...", "--"] + paths
     )
     added_lines = []
     for line in diff.splitlines():