Multiple issues around the new function of -max_error_rate
|Reported by:||JayBlanc||Owned by:|
|Blocking:||Reproduced by developer:||no|
|Analyzed by developer:||no|
The new function -max_error_rate provides a way of issuing a failure return value if there are too many errors encountered during operation.
However the implementation has the following flaws,
- It is an undocumented change into the basic function of the CLI tool, and as such breaks many scripted and embedded uses of ffmpeg that do not expect positive return values from non-fatal errors.
- The 'percentage' does not seem to be very well quantified, as to if this means recoverable or non-recoverable errors. Media that decodes perfectly, but has packets that had errors, can trigger this. ie, poorly encoded but functionally decodable media, such as a flash svc2 file from a third party encoder that is decoded correctly but has repeated errors. Or streams that have redundancy or built in error correction, such as Over The Air TS streams.
My recommended change would be to make it default to a value of 1, and for errors to only be counted towards the percentage level if they are non-recoverable.
Change History (12)
comment:9 by , 8 years ago
|Component:||undetermined → ffmpeg|
|Priority:||normal → important|
|Status:||new → open|