GitLab

BUG: 437854
FIXED-IN: 21.04.2

For multi-leg journeys all involved trains are listed here, but neither
locations nor times of the involved connections. Since we cannot identify
all elements of journey correctly then, don't pretend we do that and ignore
the train number in that case. This also avoids merge problems when more
details become available later by other means.

Return journeys were not handled at all yet, we can use the same logic for
that with just different XPath search patterns.

BUG: 436976
FIXED-IN: 21.04.2

The problem here is that the way we extracted the date from a JS Date
object was prone to timezone conversion rounding errors, possibly giving
us a day before or after as well. We fix this by letting the post-processor
deal with the problem for us.

This makes post-processing handle the case that the departure time only
contains a time by using the departure day value to fill the gap.

This avoids extractor scripts having to do error prone date/time
computations in JS, in particular if both values are extracted from
different document nodes (e.g. a barcode and a PDF body).

Fixes the Craft build for 21.04 or higher.

This will simplify upcoming changes to these files.

This is a first step towards cleaning up some data type specific
complexity in the filter loading code.

GIT_SILENT: add sanitizers.supp so we can suppress some leak generated by some lib (as libfontconfig)

(cherry picked from commit b8ee963b)

We can meanwhile extract that reliably from the U_HEAD block of the
UIC 918.3 data.

Replaces manual decoding in the Thalys extractor.

Down to only two missing files for full REUSE compliance now.

This now all uses the new decoding utilities that do proper boundary
checks, and reduces the amount of duplicated magic numbers. This also
found two issues in the existing boundary checks, and a wrong deciding
of the RCT2 field format flag.

Otherwise it fails to build because the assignment operator for char is deleted

This adds support for version 2 according to ERA TAP TSI TD B.12 with its
enlarged signature field.

This also factors out the low-level decoding utility functions as we need
them independent of the Uic9183Block structure.

Those contain a more reliable validity time range for return tickets,
compared to all other date/time sources in here.

Those contain local public transport extensions for DB long distance
tickets, using elements of the VDV ticket format.

This allows us to print almost all content of a VDV ticket in the barcode
dumping tool, as well as the removal of a lot of diagnostic code from the
actual VDV ticket parser.

The validity area structs still need work to properly decode those, but I
yet have to find the official documents describing those.  We should have
enough information to decode 0080VU blocks in UIC 918.3 tickets now though.

Also add SPDX annotations to the build system files while at it.

Most importantly this now makes the endian conversion transparent, even
for the unusual 3 byte numeric types we couldn't handle before. This also
adds implicit conversion for the date and date/time types.