Re: New Git hooks for checking file names in commit messages

[Jim Porter <jporterbugs@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 967 bytes --]

On 4/22/2023 12:51 AM, Eli Zaretskii wrote:
> The programmatic expression of the standards is in add-log.el, I
> think.

Thanks, that was helpful. From that, I see that you can list multiple 
files separated by commas and a space like so:

   * file1.txt, file2.txt

add-log.el's font-lock doesn't allow lists of files like that to span 
multiple lines, but I think it's reasonable to support that in the 
hooks. Everything else about the entries can span multiple lines, after all.

With add-log.el as a reference, I tightened up the Git hooks a bit so 
that they're stricter. They're still reasonably flexible though, e.g. it 
won't treat something like this as a file entry:

   * This is a bullet point.

That will help reduce false positives so that committers can format the 
description of their commit fairly freely. Hopefully this is a good 
balance: the hooks are stricter than authors.el, but not so strict that 
reasonable (IMO) messages will be rejected.

[-- Attachment #2: 0001-Improve-the-logic-of-the-file-entry-Git-hooks-to-mat.patch --]
[-- Type: text/plain, Size: 2617 bytes --]

From f86ddf1ced484f673eada070f35da7d052e43b99 Mon Sep 17 00:00:00 2001
From: Jim Porter <jporterbugs@gmail.com>
Date: Fri, 21 Apr 2023 10:06:49 -0700
Subject: [PATCH] Improve the logic of the file entry Git hooks to match
 admin/authors.el

* build-aux/git-hooks/commit-msg-files.awk (check_commit_msg_files):
Accumulate file entries over multiple lines.
---
 build-aux/git-hooks/commit-msg-files.awk | 31 ++++++++++++++++++------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/build-aux/git-hooks/commit-msg-files.awk b/build-aux/git-hooks/commit-msg-files.awk
index 3856e474d3e..edffb7d339c 100644
--- a/build-aux/git-hooks/commit-msg-files.awk
+++ b/build-aux/git-hooks/commit-msg-files.awk
@@ -59,15 +59,28 @@ function check_commit_msg_files(commit_sha, verbose,    changes, good, \
     if (verbose && ! msg)
       msg = $0
 
-    # Find lines that reference files.  We look at any line starting
-    # with "*" (possibly prefixed by "; ") where the file part starts
-    # with an alphanumeric character.  The file part ends if we
-    # encounter any of the following characters: [ ( < { :
-    if (/^(; )?\*[ \t]+[[:alnum:]]/ && match($0, /[[:alnum:]][^[(<{:]*/)) {
-      # There might be multiple files listed on this line, separated
+    # Find file entries in the commit message.  We look at any line
+    # starting with "*" (possibly prefixed by "; ") followed by a ":",
+    # possibly on a different line.  If we encounter a blank line
+    # without seeing a ":", then we don't treat that as a file entry.
+
+    # Accumulate the contents of a (possible) file entry.
+    if (/^[ \t]*$/)
+      filenames_str = ""
+    else if (/^(; )?\*[ \t]+[[:alnum:]]/)
+      filenames_str = $0
+    else if (filenames_str)
+      filenames_str = (filenames_str $0)
+
+    # We have a file entry; analyze it.
+    if (filenames_str && /:/) {
+      # Delete the leading "*" and any trailing information.
+      sub(/^(; )?\*[ \t]+/, "", filenames_str)
+      sub(/[[(<:].*$/, "", filenames_str)
+
+      # There might be multiple files listed in this entry, separated
       # by spaces (and possibly a comma).  Iterate over each of them.
-      split(substr($0, RSTART, RLENGTH), filenames, ",?([[:blank:]]+|$)")
-
+      split(filenames_str, filenames, ",[ \t]+")
       for (i in filenames) {
         # Remove trailing slashes from any directory entries.
         sub(/\/$/, "", filenames[i])
@@ -83,6 +96,8 @@ function check_commit_msg_files(commit_sha, verbose,    changes, good, \
           good = 0
         }
       }
+
+      filenames_str = ""
     }
   }
   close(cmd)
-- 
2.25.1