On 08/10/18 10:22, I wrote:
>
> That is not quite the whole story. There were a few programmes whose
> subtitle XML files (TTML?) were corrupt. They included some NUL
> characters and, I am told, NUL is not a valid character in XML.
>
> The code was changed to ignore NUL characters. As an additional measure
> downloading of subtitles was delayed until the end to reduce the risk
> that errors in subtitles could cause the whole download to fail.
>
I haven't got it quite right either. The 3.07 release notes say,
"Fixed a bug that caused get_iplayer to exit (and skip metadata tagging)
when corrupt subtitles were provided by the BBC. Corrupt subtitles will
now generate an error message, but get_iplayer will not exit. As an
extra precaution, subtitles download now occurs after tagging is complete."
They also say,
"Subtitles are now trimmed according to --start and --stop, if
specified. The actual subtitles start and/or stop times are calculated
to the nearest chunk boundary in the HLS or DASH stream. The value of
--suboffset is still applied. This trimming is not supported for
--subtitles-only. You can use --subtitles-only to re-download the full
subtitles for a programme if desired. (@notnac)"
I am guessing, but it is possible that a change in download order was
needed to implement that.
The 3.09 release notes say,
"Null characters are now removed from subtitles before processing. This
should prevent at least some problems caused by corrupt subtitle files.
(@notnac)"
Best wishes
Richard