[PATCH v3 4/4] parse_commit(): describe more date-parsing failure modes

[Jeff King <peff@peff.net>]

The previous few commits improved the parsing of dates in malformed
commit objects. But there's one big case left implicit: we may still
feed garbage to parse_timestamp(). This is preferable to trying to be
more strict, but let's document the thinking in a comment.

Signed-off-by: Jeff King <peff@peff.net>
---
 commit.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/commit.c b/commit.c
index 8dfe92cf37..e2e4fd2db9 100644
--- a/commit.c
+++ b/commit.c
@@ -148,6 +148,15 @@ static timestamp_t parse_commit_date(const char *buf, const char *tail)
 	/*
 	 * We know there is at least one digit (or dash), so we'll begin
 	 * parsing there and stop at worst case at eol.
+	 *
+	 * Note that we may feed parse_timestamp() extra characters here if the
+	 * commit is malformed, and it will parse as far as it can. For
+	 * example, "123foo456" would return "123". That might be questionable
+	 * (versus returning "0"), but it would help in a hypothetical case
+	 * like "123456+0100", where the whitespace from the timezone is
+	 * missing. Since such syntactic errors may be baked into history and
+	 * hard to correct now, let's err on trying to make our best guess
+	 * here, rather than insist on perfect syntax.
 	 */
 	return parse_timestamp(dateptr, NULL, 10);
 }
-- 
2.40.1.663.g410c33770c.dirty