As the $rootpath is concatenated with the $relpath, we would end with a
path like course_files//file.txt. Such a path is detected as not-valid
later and an error would be thrown.
So the patch makes sure that the trailing slash is cut off.
The patch introduces two new tests for moodle1_file_manager::migrate_directory().
The second test with the trailing slash reveals the trouble here.
The expected behaviour is that the method would deal with the trailing
slash. Debugging message for the developer should be displayed though.
URL of files referenced in HTML fields via $@FILEPHP$@ are usually
encoded if they were linked using the HTML editor. So if they contain
spaces or plus signs in the name, their URL is encoded to use %20 and
%2B respectively.
Sometimes, for example when the HTML was edited manually in 1.9 without
the editor, non-encoded file names may be put into the HTML text. Such a
link used to work in 1.9 so we add an explicit support for it, too.
However, we do not support partially encoded URLs (e.g. only spaces are
encoded while plus signs are not). Such links are unmodified and will
rely on Legacy files support.
This handler will automatically be used for any block which does not provide its own handler, and should be subclassed by any block which needs to do anything extra (or different).
This commit incorporates contributions from Mike Churchward and API design suggestions from David Mudrak.
If filenames are not run through urldecode(), those that contain spaces (or other url-encoded characters) will not be handled, as they can't be found in the backup file.
This reduces the period of time that some of the unit tests
for calculate_next_automated_backup() are failing right now
(from 1week to just some hours around the DST changes).
Note it's not the final solution because they will still fail
when New_York or Brussels DST changes are about to happen, but
to get this completely fixed we'll need to wait to switch from
current custom TZ/DST support to PHP built-in one.
On backup, contents are cleaned to be safe utf8. That was leading to
some matching problems on restore against not cleaned DB contents (for
already existing questions). This commit adds one fallback method to
perform the match against cleaned DB contents.
