CNA-BUG-144: VideoPlayer::GetTexture lets the decoder's std::runtime_error escape, undocumented, and leaves the player Playing

CNA snapshot 009d40f5  ·  Known Issues › Current bugs  ·  source links pinned to 009d40f5

A decode, packet, resample or I/O error mid-stream is thrown by VideoDecoder::NextFrame as std::runtime_error through GetTexture, whose header documents only ObjectDisposedException, and the player stays Playing so the next call tries again.

Identifier

CNA-BUG-144

Category

Bug

Subsystem

Audio & media

Status

Open

Verified against

CNA 009d40f5 (009d40f5dd085c4e674d3479675fac84b12b3e0a)

Severity

Low (a triage suggestion, not a project priority)

Evidence basis

Source-verified: read at TARGET, not executed

Tests touching this area

Yes: see Current tests

Affected contract

VideoPlayer::GetTexture() exceptions and the player state after a mid-stream decode failure

Expected behaviour

VideoPlayer.hpp documents GetTexture as throwing ObjectDisposedException only. Play treats a decoder failure transactionally: OpenDecoder catches, closes the decoder, resets the state to Stopped and rethrows. The frame path should either be documented as throwing, with a defined state afterwards, or report the failure through the state.

Actual behaviour at TARGET

GetTexture in VideoPlayer.cpp loops over decoder_->NextFrame and DrainAndFlushAudioBuffer with no try. NextFrame and ProcessAudioPacket in VideoDecoder.cpp throw std::runtime_error for a video decode error, a rejected or failed packet send, a failed flush, a demux I/O error and audio decode or resample failures (a deliberate change so that corruption is no longer mistaken for end of stream). The exception reaches the caller, usually inside Draw. It is not a System::Exception, so a handler written for that type misses it. The player stays Playing with the decoder open, and every later GetTexture re-enters the decoder.

Source locations

• modules/media/src/Xna/Video/VideoPlayer.cpp — VideoPlayer::GetTexture (no try) versus OpenDecoder (catch, CloseDecoder, rethrow)
• modules/video-ffmpeg/src/VideoDecoder.cpp — VideoDecoder::NextFrame and ProcessAudioPacket throw sites
• modules/media/include/Microsoft/Xna/Framework/Media/Video/VideoPlayer.hpp — GetTexture documentation
• modules/media/tests/CNA/Internal/Media/VideoDecoderTests.cpp — CorruptedMidStreamDataThrowsRatherThanSilentlyEndingCleanly

Evidence

Checked by reading at 009d40f5; not executed. The decoder-level throw is pinned by VideoDecoderTest.CorruptedMidStreamDataThrowsRatherThanSilentlyEndingCleanly; what the player does with it follows from GetTexture's code.

Focused reproduction

No focused reproduction is known. Nothing has been invented here; the evidence above is what exists.

Current tests

VideoDecoderTests.cpp pins the throw at decoder level with corrupt_test_h264.mp4. VideoPlayerTests.cpp has no case that plays a corrupt file through GetTexture.

Regression test

Play corrupt_test_h264.mp4 through a VideoPlayer and call GetTexture until the corruption is reached; require the documented outcome (for example an exception of a documented type followed by Stopped) and no second throw from the next call.

Blast radius

Builds with the FFmpeg backend playing corrupt or truncated-in-the-middle files. Clean files and trailing truncation of Matroska files (reported as end of stream) are unaffected.

Workaround

Catch std::exception around GetTexture and call Stop() on failure.