Thread (8 messages) 8 messages, 4 authors, 2025-03-25

Re: [PATCH] blame: fix unblamable and ignored lines in porcelain mode

From: Patrick Steinhardt <hidden>
Date: 2025-03-24 10:16:36
Subsystem: the rest · Maintainer: Linus Torvalds

Possibly related (same subject, not in this thread)

On Sun, Mar 23, 2025 at 08:58:03AM -0700, Junio C Hamano wrote:
Karthik Nayak [off-list ref] writes:
quoted
However, this option was never extended to the porcelain mode of
'git-blame(1)'. Since the documentation does not indicate this
exclusion, it is a bug.
I agree it is a bug when people added ignore or unblamable support
that they did not _consider_ what to do with their new pieces of
information to help porcelain writers.  It is not a bug in the code
per-se, but it is a bug in the brain of these people ;-)

But prefixing random garbage to the commit object name line in the
porcelain mode output does not sound like the right solution to the
bug, either.

When enhancing an existing output format, make sure that your
changes will have minimum empact to existing parsers that do not
know about your extension.  It is reasonably expected that existing
Porcelain scripts reading from --porcelain mode output works by

 - Recognizing a line that match "^[0-9a-f]{40} \d+ \d+ \d+$" and
   take it as the beginning of a new record;

 - Collect all info lines before the payload line.  Lines that
   describe per-commit information are not repeated if it is already
   shown, so remember them when you see the commit for the first
   time, and recall them when you recognize the commit you already
   saw.

 - A payload line is indented with HT and terminates the record.

If you start to add unrecognizable garbage to the line with very
well known fixed format that is used as record delimiter, you would
break the existing parsers, which is not a very nice thing to do.
Are there other and better ways you can think of to add new pieces
of information like this in a way with less severe damage?
I think the porcelain mode is already built so that it can be extended
with arbitrary new information, no? In `emit_one_suspect_detail()` we
end up printing one line per info we want to display. I would have
expected that we can extend that function to also print information
around unblamable or ignored commits, like we already do for boundary
commits. E.g. something like the patch further down.

Thanks!

Patrick
diff --git a/builtin/blame.c b/builtin/blame.c
index c470654c7ec..cd8322e2619 100644
--- a/builtin/blame.c
+++ b/builtin/blame.c
@@ -255,7 +255,8 @@ static void write_filename_info(struct blame_origin *suspect)
  * the first time each commit appears in the output (unless the
  * user has specifically asked for us to repeat).
  */
-static int emit_one_suspect_detail(struct blame_origin *suspect, int repeat)
+static int emit_one_suspect_detail(struct blame_entry *ent,
+				   struct blame_origin *suspect, int repeat)
 {
 	struct commit_info ci = COMMIT_INFO_INIT;
 
@@ -275,6 +276,10 @@ static int emit_one_suspect_detail(struct blame_origin *suspect, int repeat)
 	printf("summary %s\n", ci.summary.buf);
 	if (suspect->commit->object.flags & UNINTERESTING)
 		printf("boundary\n");
+	if (mark_unblamable_lines && ent->unblamable)
+		printf("unblamable\n");
+	if (mark_ignored_lines && ent->ignored)
+		printf("ignored\n");
 
 	commit_info_destroy(&ci);
 
@@ -295,7 +300,7 @@ static void found_guilty_entry(struct blame_entry *ent, void *data)
 		printf("%s %d %d %d\n",
 		       oid_to_hex(&suspect->commit->object.oid),
 		       ent->s_lno + 1, ent->lno + 1, ent->num_lines);
-		emit_one_suspect_detail(suspect, 0);
+		emit_one_suspect_detail(ent, suspect, 0);
 		write_filename_info(suspect);
 		maybe_flush_or_die(stdout, "stdout");
 	}
@@ -344,9 +349,10 @@ static const char *format_time(timestamp_t time, const char *tz_str,
 #define OUTPUT_COLOR_LINE           (1U<<10)
 #define OUTPUT_SHOW_AGE_WITH_COLOR  (1U<<11)
 
-static void emit_porcelain_details(struct blame_origin *suspect, int repeat)
+static void emit_porcelain_details(struct blame_entry *ent,
+				   struct blame_origin *suspect, int repeat)
 {
-	if (emit_one_suspect_detail(suspect, repeat) ||
+	if (emit_one_suspect_detail(ent, suspect, repeat) ||
 	    (suspect->commit->object.flags & MORE_THAN_ONE_PATH))
 		write_filename_info(suspect);
 }
@@ -366,7 +372,7 @@ static void emit_porcelain(struct blame_scoreboard *sb, struct blame_entry *ent,
 	       ent->s_lno + 1,
 	       ent->lno + 1,
 	       ent->num_lines);
-	emit_porcelain_details(suspect, repeat);
+	emit_porcelain_details(ent, suspect, repeat);
 
 	cp = blame_nth_line(sb, ent->lno);
 	for (cnt = 0; cnt < ent->num_lines; cnt++) {
@@ -376,7 +382,7 @@ static void emit_porcelain(struct blame_scoreboard *sb, struct blame_entry *ent,
 			       ent->s_lno + 1 + cnt,
 			       ent->lno + 1 + cnt);
 			if (repeat)
-				emit_porcelain_details(suspect, 1);
+				emit_porcelain_details(ent, suspect, 1);
 		}
 		putchar('\t');
 		do {
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help