Bug 232712

Summary:

[MSE][GStreamer] Use full nanosecond precision timestamps

Product:

WebKit

Reporter:

Alicia Boya García <aboya>

Component:

WebKitGTK

Assignee:

Alicia Boya García <aboya>

Status:

NEW ---

Severity:

Normal

CC:

bugs-noreply, calvaris, cgarcia, eric.carlson, ews-watchlist, glenn, gustavo, jer.noble, menard, philipj, philn, pnormand, sergio, vjaquez

Priority:

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	calvaris: review+

Description Alicia Boya García 2021-11-04 08:00:07 PDT

This patch modifies MediaSampleGStreamer to use full nanosecond
precision instead of microsecond precision.

This fixes media/media-source/media-mp4-h264-sequence-mode.html

Comment 1 Alicia Boya García 2021-11-04 08:12:57 PDT

Created attachment 443304 [details]
Patch

Comment 2 Alicia Boya García 2021-11-12 07:10:07 PST

Comment on attachment 443304 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=443304&action=review

> Source/WebCore/ChangeLog:11
> +        This fixes media/media-source/media-mp4-h264-sequence-mode.html

The fix may be accidental, since removing h264 parse from the pipeline broke the test again. In both cases the failure is from a minute discrepancy in timestamps.

Comment 3 Philippe Normand 2022-04-29 03:49:29 PDT

What's the status of this patch? It was reviewed but never merged.

Comment 4 Philippe Normand 2022-04-29 08:54:16 PDT

Maybe the test was updated since then, but with this patch (rebased), it still fails:

--- /app/webkit/WebKitBuild/Release/layout-test-results/media/media-source/media-mp4-h264-sequence-mode-expected.txt
+++ /app/webkit/WebKitBuild/Release/layout-test-results/media/media-source/media-mp4-h264-sequence-mode-actual.txt
@@ -10,13 +10,13 @@
 EVENT(update)
 EXPECTED (sourceBuffer.timestampOffset == '0') OK
 EXPECTED (sourceBuffer.buffered.length == '1') OK
-EXPECTED (sourceBuffer.buffered.start(0) == '1') OK
-EXPECTED (sourceBuffer.buffered.end(0) == '2') OK
+EXPECTED (sourceBuffer.buffered.start(0) == '1'), OBSERVED '1.041666666' FAIL
+EXPECTED (sourceBuffer.buffered.end(0) == '2'), OBSERVED '2.041666666' FAIL
 RUN(sourceBuffer.mode="sequence")
 RUN(sourceBuffer.appendBuffer(loader.mediaSegment(1)))
 EVENT(update)
 EXPECTED (sourceBuffer.timestampOffset == '1') OK
 EXPECTED (sourceBuffer.buffered.length == '1') OK
-EXPECTED (sourceBuffer.buffered.end(0) == '3') OK
+EXPECTED (sourceBuffer.buffered.end(0) == '3'), OBSERVED '3.041666666' FAIL
 END OF TEST

Comment 5 Philippe Normand 2022-04-29 08:58:55 PDT

With current ToT:

--- /app/webkit/WebKitBuild/Release/layout-test-results/media/media-source/media-mp4-h264-sequence-mode-expected.txt
+++ /app/webkit/WebKitBuild/Release/layout-test-results/media/media-source/media-mp4-h264-sequence-mode-actual.txt
@@ -10,13 +10,13 @@
 EVENT(update)
 EXPECTED (sourceBuffer.timestampOffset == '0') OK
 EXPECTED (sourceBuffer.buffered.length == '1') OK
-EXPECTED (sourceBuffer.buffered.start(0) == '1') OK
-EXPECTED (sourceBuffer.buffered.end(0) == '2') OK
+EXPECTED (sourceBuffer.buffered.start(0) == '1'), OBSERVED '1.041666' FAIL
+EXPECTED (sourceBuffer.buffered.end(0) == '2'), OBSERVED '2.041666' FAIL
 RUN(sourceBuffer.mode="sequence")
 RUN(sourceBuffer.appendBuffer(loader.mediaSegment(1)))
 EVENT(update)
 EXPECTED (sourceBuffer.timestampOffset == '1') OK
 EXPECTED (sourceBuffer.buffered.length == '1') OK
-EXPECTED (sourceBuffer.buffered.end(0) == '3') OK
+EXPECTED (sourceBuffer.buffered.end(0) == '3'), OBSERVED '3.041666' FAIL
 END OF TEST

Comment 6 Philippe Normand 2022-08-16 11:56:41 PDT

(In reply to Philippe Normand from comment #3)
> What's the status of this patch? It was reviewed but never merged.

Ping Alicia?

Comment 7 Alicia Boya García 2022-08-19 05:01:59 PDT

(In reply to Philippe Normand from comment #6)
> (In reply to Philippe Normand from comment #3)
> > What's the status of this patch? It was reviewed but never merged.
> 
> Ping Alicia?

I am aware of it, just had my attention on other things. My biggest problem with this patch was tests regressing in convoluted ways that should hopefully have been fixed by previous patches. Nowadays it would need a rebase and re-test.