On Sat, 12 Apr 2025 01:43:29 GMT, Alexander Matveev <almat...@openjdk.org>
wrote:
> - Fixed by reloading decoder for each seek.
> - Tested with all H.265 files for HLS/HTTP/FILE, no issues found.
> - Seek performance is not affected or at least I did not notice any
> performance issues when doing reload for each seek.
>
> This is workaround and no other reasonable solutions were found.
I left a few questions on the code changes. I haven't tested it yet.
modules/javafx.media/src/main/native/gstreamer/plugins/mfwrapper/mfwrapper.cpp
line 259:
> 257: // pDecoderOutput is released.
> 258: SafeRelease(&decoder->pDecoderOutput);
> 259: decoder->pDecoderBuffer = NULL;
Is setting this to NULL sufficient or does it need to be freed / released first?
modules/javafx.media/src/main/native/gstreamer/plugins/mfwrapper/mfwrapper.cpp
line 267:
> 265: // pColorConvertOutput is released.
> 266: SafeRelease(&decoder->pColorConvertOutput[i]);
> 267: decoder->pColorConvertBuffer[i] = NULL;
Is setting this to NULL sufficient or does it need to be freed / released first?
modules/javafx.media/src/main/native/gstreamer/plugins/mfwrapper/mfwrapper.cpp
line 1107:
> 1105: {
> 1106: SafeRelease(&decoder->pColorConvertOutput[i]);
> 1107: decoder->pColorConvertBuffer[i] = NULL;
Same question as above.
modules/javafx.media/src/main/native/gstreamer/plugins/mfwrapper/mfwrapper.cpp
line 1574:
> 1572: g_strdup("Failed to reload
> decoder"), NULL,
> 1573: ("mfwrapper.c"),
> ("mfwrapper_sink_event"), 0);
> 1574: }
Is this a fatal error? If so, do you need to return early here?
-------------
PR Review: https://git.openjdk.org/jfx/pull/1775#pullrequestreview-2776106535
PR Review Comment: https://git.openjdk.org/jfx/pull/1775#discussion_r2049156892
PR Review Comment: https://git.openjdk.org/jfx/pull/1775#discussion_r2049157580
PR Review Comment: https://git.openjdk.org/jfx/pull/1775#discussion_r2049164737
PR Review Comment: https://git.openjdk.org/jfx/pull/1775#discussion_r2049172358
