VTM9.0 decoder crash when using debug mode

[chiaming]

When using the following encoding options, even with the bugfix in #1112, decoder will crash in HLSyntaxReader::parseSliceHeader(...) . But release mode has no such issue.

EncoderApp.exe -c ..\cfg\encoder_randomaccess_vtm.cfg -c ..\cfg\per-sequence\RaceHorses.cfg -f 3

And after enabling ENABLE_TRACING, as the following partial trace logs from encoder/decoser side, some flags are unexpectedly parsed and configured. My IDE is Visual Studio 2017.

Partial trace log at encoder side:
=========== Sequence Parameter Set ===========
...
sps_ladf_enabled_flag u(1) : 0
sps_scaling_list_enabled_flag u(1) : 0
sps_scaling_matrix_designated_colour_space_flag u(1) : 1
sps_dep_quant_enabled_flag u(1) : 1
sps_virtual_boundaries_enabled_flag u(1) : 0
sps_general_hrd_parameters_present_flag u(1) : 0
field_seq_flag u(1) : 0
vui_parameters_present_flag u(1) : 0
sps_extension_present_flag u(1) : 0
rbsp_stop_one_bit u(1) : 1
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
...

Partial trace log at decoder side:
=========== Sequence Parameter Set ===========
...
sps_ladf_enabled_flag u(1) : 0
sps_explicit_scaling_list_enabled_flag u(1) : 0
sps_dep_quant_enabled_flag u(1) : 1
sps_virtual_boundaries_enabled_flag u(1) : 1
sps_loop_filter_across_virtual_boundaries_present_flag u(1) : 0
sps_general_hrd_params_present_flag u(1) : 0
field_seq_flag u(1) : 0
vui_parameters_present_flag u(1) : 0
sps_extension_present_flag u(1) : 0
rbsp_stop_one_bit u(1) : 1
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
rbsp_alignment_zero_bit u(1) : 0
...

[ksuehring]

It seems possible that the encoder sets sps_scaling_matrix_for_alternative_colour_space_disabled_flag to 1 based on a config file setting, although it is not present in the bitstream. But this would cause sps_scaling_matrix_designated_colour_space_flag to be written to the bitstream at the encoder side. The decoder would then use the inferred value of 0 for sps_scaling_matrix_for_alternative_colour_space_disabled_flag and subsequently not read sps_scaling_matrix_designated_colour_space_flag.

I'm attaching a patch. Can you check, if it solves your problem?

[chiaming]

Yes, after applying the provided patch, the crash issue is gone and no mismatch. Thank you.

[ksuehring]

I submitted as

​https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM/-/merge_requests/1680

[ksuehring]

The MR was merged