| Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
MXF::set_key() sets it.
|
|
required mangling of long filenames on Windows.
Also wrap lots of missing places (e.g. calls to asdcplib, libxml++,
libcxml etc.) in dcp::filesystem::fix_long_path().
The idea is to keep paths un-mangled until they we call some
filesystem-related API and mangle them at that point. Otherwise
we end up serialising mangled names, which seems like it will
not end well.
Should fix DoM #2623.
|
|
|
|
|
|
|
|
This exposes a useless assignment of export_differing_subtitles,
which I've removed (and altered the comment accordingly).
|
|
|
|
This is simpler and allows us to remove the hack of allocating some
"maximum" buffer for incoming JPEG2000 data.
It does mean that the buffer is zero-initialized before being written
to, but hopefully that doesn't matter too much.
|
|
and default to writing the 2014 namespace.
|
|
they say that subtitle <IssueDate> must be of the form yyyy-mm-ddThh:mm:ss
They don't explain it, and the standards don't say this (as far as I can see)
but as usual their QC fails <IssueDates> that don't meet this "spec" so here
we are.
See also:
https://dcpomatic.com/forum/viewtopic.php?t=2069
|
|
|
|
This is an attempt to fix DoM bug #2205.
|
|
In DoM bug #2061 it is reported that Easy DCP gives the error
"XML Document has default root namespace prefix: dcst.
Default namespace should not use prefix for root or root namespace child nodes."
with SMPTE subtitle files written by DCP-o-matic, and that the correct
fix is to make the former dcst: namespace the default and then remove
the dcst: qualifiers from everything.
I'm not sure I agree with the error; AFAICS the subtitle files written
by previous versions did not have a default root namespace prefix,
since it was specified by xmlns:dcst and not just xmlns alone, so I
think they were valid.
However, using a default NS also seems fine, slightly simplifies the
code and produces more compact subtitle files, so we change that here.
It should also stop Easy DCP complaining, which is always marginally
preferable to sticking to our guns and getting the blame for it.
|
|
This is also unavailable if the asset is encrypted.
|
|
Before this if we tried to get the XML of an encrypted asset we
would just get an empty string. Now we get a boost::none which
means the verifier can avoid trying to check details of the XML
(and instead raise a warning that you are trying to verify data
that it cannot decrypt).
|
|
|
|
|
|
They should not be the same, so it's unhelpful to make them so.
parse_xml() sets up _xml_id anyway.
|
|
|
|
Check that subtitles don't overlap reel boundaries, and fix a few
tests that trip this check.
Fix confusion when calculating subtitle timings during verification
where the picture asset frame rate was being used rather than the
subtitle asset's edit rate.
Do the subtitle timing verification for Interop as well as SMPTE
subtitles.
Take <StartTime> tags into account when checking subtitles, even
though Bv2.1 says they should be set to 0.
Rename Time::as_editable_units to Time::as_editable_units_ceil
and add a _floor variant, then use that to round down when checking
reel boundary overlaps.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
subtitle assets so that it cab be verified without any
interference from being passed through libdcp.
|
|
|
|
writing lots of subs (DCP-o-matic bug #1561).
|
|
|
|
If we load SMPTE subtitles from a XML file try to load PNG data
from the same directory; this feels like a hack.
|
|
|
|
instead of the namespace URI for what asdcplib calls
TimedTextDescriptor::NamespaceName.
Thanks to clairmeta; see https://github.com/Ymagis/ClairMeta/issues/72
|
|
be the same as the internal XML ID. This in turn must be different
from the MXF's UUID.
I *think* that we have:
AssetUUID: MXF's UUID, which must be different; written from
the WriterInfo (set up by fill_writer_info, which is passed _id)
AssetID: MXF's ResourceID, which must be the _xml_id; written from
the TimedTextDescriptor.
I believe that this patch fixes the ResourceID to be the XML ID (as
required) without undoing the Doremi fix whereby the MXF UUID must
be different to the XML ID.
This patch quells ClairMeta's error.
|
|
|
|
|
|
reading image subtitles.
|
|
indices). Basic and untested SMPTE image subtitle support.
|
|
|
|
|
|
|