Changes between Initial Version and Version 2 of Ticket #8487
- Timestamp:
- Jan 19, 2020, 12:58:42 PM (7 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #8487
- Property Keywords adpcm added
- Property Priority normal → minor
- Property Reproduced by developer set
- Property Status new → open
- Property Version 4.2 → git-master
-
Ticket #8487 – Description
initial v2 3 3 Output: 4 4 ----------------------------------------------------------------------------- 5 {{{ 5 6 ffmpeg version 4.2.2 Copyright (c) 2000-2019 the FFmpeg developers 6 7 built with clang version 8.0.0-3~ubuntu18.04.2 (tags/RELEASE_800/final) … … 26 27 Stream #0:0: Audio: adpcm_ea_xas, mono, s16p 27 28 At least one output file must be specified 28 ----------------------------------------------------------------------------- 29 }}} 29 30 30 31 Look at: 31 32 {{{ 32 33 case AV_CODEC_ID_ADPCM_EA_XAS: 33 34 for (channel=0; channel<avctx->channels; channel++) { … … 48 49 } 49 50 50 51 }}} 51 52 This bug really does not make sense to me, val is controlled but the 0x0f mask should prevent to have a negative shift[n] value. 52 53 UBSan has no false positives, so there is a bug indeed, but I don't understand why. … … 54 55 55 56 I attach a file to reproduce the issue in base64: 57 {{{ 56 58 BAAAAABQX0UvBf//Bf93/3wIAAAAKAAAABBFMDAAABAATjs0MP93/1NBVUNFKTBtTncIAAAAE15BVV9FLzAMRm7/d/93B+4AACgAAABiRTAlOzQwbU47NCBtTgA= 59 }}}


