threads / patch / 51220

patchblame - fix some issues identified by coverage report.

Subject: [PATCH] blame - fix some issues identified by coverage report.

## tl;dr

3 messages between Jun 1, 2019 and Jun 3, 2019. Diffs are folded; open one to read it.

replies: 2people: 3as markdown or json

michael@platin.gs· Jun 1, 2019, 21:09 UTC · lore
From: Michael Platings <michael@platin.gs>
Thanks to Derrick Stolee for highlighting missing coverage.

In the case of "certainties[i] = CERTAINTY_NOT_CALCULATED" this was defeating an optimization that preserved results of calculations between line-matching passes. This had caused other code to never be executed - that code is now executed and only discards calculation results that are no longer valid.

In the case of "max_search_distance_b = 0" this was never executed because another statement was added earlier in the function to return early in such a case.

Signed-off-by: Michael Platings <michael@platin.gs>
---
 blame.c | 12 ++----------
 1 file changed, 2 insertions(+), 10 deletions(-)
Show changes to blame.c +2 −10
diff --git a/blame.c b/blame.c
index c059063688..2b47c942db 100644
--- a/blame.c
+++ b/blame.c
@@ -825,10 +825,6 @@ static void fuzzy_find_matching_lines_recurse(
 		*get_similarity(similarities, most_certain_line_a - start_a,
 				i, closest_local_line_a,
 				max_search_distance_a) = -1;
-
-		if (certainties[i] >= 0) {
-			certainties[i] = CERTAINTY_NOT_CALCULATED;
-		}
 	}
 
 	/* More invalidating of results that may be affected by the choice of
@@ -947,12 +943,8 @@ static int *fuzzy_find_matching_lines(struct blame_origin *parent,
 	if (max_search_distance_a >= length_a)
 		max_search_distance_a = length_a ? length_a - 1 : 0;
 
-	if (length_a == 0) {
-		max_search_distance_b = 0;
-	} else {
-		max_search_distance_b = ((2 * max_search_distance_a + 1) *
-			length_b - 1) / length_a;
-	}
+	max_search_distance_b = ((2 * max_search_distance_a + 1) *
+		length_b - 1) / length_a;
 
 	result = xcalloc(sizeof(int), length_b);
 	second_best_result = xcalloc(sizeof(int), length_b);
-- 
2.21.0
Barret Rhoden· Jun 3, 2019, 15:26 UTC · re: michael@platin.gs · lore

Re: [PATCH] blame - fix some issues identified by coverage report.

Hi Michael -
On 6/1/19 5:09 PM, michael@platin.gs wrote:
Show 15 quoted lines
> From: Michael Platings <michael@platin.gs>
> 
> Thanks to Derrick Stolee for highlighting missing coverage.
> 
> In the case of "certainties[i] = CERTAINTY_NOT_CALCULATED" this was
> defeating an optimization that preserved results of calculations
> between line-matching passes. This had caused other code to never
> be executed - that code is now executed and only discards calculation
> results that are no longer valid.
> 
> In the case of "max_search_distance_b = 0" this was never executed
> because another statement was added earlier in the function to return
> early in such a case.
> 
> Signed-off-by: Michael Platings <michael@platin.gs>

If it's OK with you, I can squash this into your existing patch in the blame-ignore patch set.

Thanks,
Barret
Junio C Hamano· Jun 3, 2019, 18:08 UTC · re: Barret Rhoden · lore

Re: [PATCH] blame - fix some issues identified by coverage report.

Barret Rhoden <brho@google.com> writes:
Show 21 quoted lines
> Hi Michael -
>
> On 6/1/19 5:09 PM, michael@platin.gs wrote:
>> From: Michael Platings <michael@platin.gs>
>>
>> Thanks to Derrick Stolee for highlighting missing coverage.
>>
>> In the case of "certainties[i] = CERTAINTY_NOT_CALCULATED" this was
>> defeating an optimization that preserved results of calculations
>> between line-matching passes. This had caused other code to never
>> be executed - that code is now executed and only discards calculation
>> results that are no longer valid.
>>
>> In the case of "max_search_distance_b = 0" this was never executed
>> because another statement was added earlier in the function to return
>> early in such a case.
>>
>> Signed-off-by: Michael Platings <michael@platin.gs>
>
> If it's OK with you, I can squash this into your existing patch in the
> blame-ignore patch set.

Thanks, both. That'd make my life easier ;-) by allowing me to ignore this patch for now, knowing you two are on top of the issue.

← back to recent threads