Opened 4 years ago
Closed 3 years ago
#1415 closed defect (fixed)
Decoder crashed after !1927 (JVET-S0078-Rebased: Handling NoOutputOfPriorPicFlag) when encountering a second IDR picture
Reported by: | ksuehring | Owned by: | |
---|---|---|---|
Priority: | major | Milestone: | |
Component: | VTM | Version: | VTM-11.0 |
Keywords: | Cc: | ksuehring, XiangLi, fbossen, jvet@… |
Description
When a second IDR is encountered in a bitstream xFlushOutput() is called, which deletes the just decoded IDR picture. Further decoding fails trying to access the picture.
Change history (6)
comment:1 Changed 4 years ago by ksuehring
comment:2 Changed 4 years ago by hendry197
MR!1937 has been submitted to fix this problem
comment:3 Changed 3 years ago by fbossen
Do we have a bitstream to test this?
comment:4 Changed 3 years ago by fbossen
- Version set to VTM-11.0
comment:5 Changed 3 years ago by ksuehring
If I remember correctly, the issue only appeared after the implementation of JVET_S0078_NOOUTPUTPRIORPICFLAG. We disabled the macro as a work around. MR !1937 was created to fix the issue, but caused other software crashes and was thus not merged.
it seem the macro JVET_S0078_NOOUTPUTPRIORPICFLAG was accidentally removed in 07bf3979927f70760d5c303cd9b7eaba0fb4ddf4 along with the whole handling of no_output_of_prior_pics_flag.
So the issue does not exist in current VTM anymore, but als the implementation of JVET-S0078 is missing.
So it's probably fine to close the issue. We should make sure to have JVET-S0078 on the list of missing implementations though.
I think the RAP_C_HHI and RAP_D_HHI would have triggered the crash because they contain more than one IDR pictures.
comment:6 Changed 3 years ago by fbossen
- Resolution set to fixed
- Status changed from new to closed
Updated the integration status of S0078 in the tracking sheet. Closing this since it was related to an integration that was reverted.
A bitstream for reproducing the problem can be created with the "--DecodingRefreshType=2" encoder parameter.