SHM6.1 TAppDecoder.exe aborts with the following error:

SHM software: Decoder Version [6.1 (HM-14.0)][Windows][VS 1800][32 bit]

POC 0 LId: 0 TId: 0 ( I-SLICE IDR, QP 26 ) [DT 0.541] [L0 ] [L1 ] [MD5:54ab44a520a0e349bcf91086e2842597,cdbc1d715873f6bbfaa62596a7ca3638, f7528fd91c9cad0e48fe93331cb0780e,(OK)] Assertion failed: m_fifo_idx < m_fifo->size(), file SHM6.1\source\Lib\TLibCommon/TComBitStream.h, line 191

A sample bit stream with two layers is attached here: Base layer resolution: 1920x1080, 30 fps Enhancement layer resolution: 3840x2160, 30 fps

When (erroneously) max_tid_ref_present_flag=1 and NumActiveRefLayers=0 for all layers (to envision a simulcast scenario) the decoder crashes at: READ_CODE( 3, uiCode, "max_tid_il_ref_pics_plus1[i][j]" ); vps->setMaxTidIlRefPicsPlus1(i, j, uiCode);

A config parameter check forcing max_tid_ref_present_flag=0 when NumActiveRefLayers=0 for all layers can solve this issue. However, this fixes the symptoms but not the underlying problem and I'm not sure if this will not conflict with standard behaviour for different situations.

This was found in Draft 7 (JCTVC-R1008v7). This also seems to be present in the ITU-T prepublished recommendation T-REC-H.265-201410. Because of this, the milestone and version are probably incorrectly set.

The first issue is a typo, which causes xPb to be defined twice instead of yPb.

Furthermore, the variables xP and yP, although obvious in light of other sections, are undefined there.

A simplistic solution would be to change [xy]Pb to [xy]P to fix this.

The changed text would then read:

For each 16x16 prediction block of the resampled picture with the top-left luma sample location at ( xP, yP ), where xP = xB << 4 and yP = yB << 4, for xB = 0..( ( PicWidthInSamplesCurrY + 15 ) >> 4 ) − 1 and yB = 0..( ( PicHeightInSamplesCurrY + 15 ) >> 4 ) − 1, its motion and mode parameters rsPredMode[ xP][ yP ], rsMvLX[ xP][ yP ], rsRefIdxLX[ xP][ yP ], and rsPredFlagLX[ xP][ yP ], with X being equal to 0 and 1, PicHeightInSamplesCurrYare derived by applying the following ordered steps:

1. The center location ( xPCtr, yPCtr ) of the luma prediction block is derived as follows:

xPCtr = xP + 8 (H-65) yPCtr = yP + 8 (H-66)