There are various places where it's not guaranteed that the
variable being used is array, and instead, can be null, bool, int...
We need to check that because php74 warns about it.
Where possible we have used the coalesce operator as
replacement for isset() ternary operations.
This just deletes all the upgrade steps previous to 3.5.0. Some
small adjustments, like tweaking globals can also be applied
when needed.
Also includes an upgrade step to prevent upgrading from any
version < 2018051700 (v3.5.0) as anti-cheating measure.
Previous commits have removed/deprecated all the upgradelib functions
not used anymore in codebase. Deletion has been documented in corresponding
upgrade.txt files:
- upgrade_fix_block_instance_configuration()
- upgrade_theme_is_from_family(), upgrade_find_theme_location()
and linkcoursesectionsupgradescriptwasrun setting
- upgrade_block_positions
- upgrade_fix_config_auth_plugin_names()
and upgrade_fix_config_auth_plugin_defaults()
- format_xxx_upgrade_remove_numsections(), format_xxx_upgrade_hide_extra_sections()
and format_xxx_upgrade_add_empty_sections()
- filter_mathjaxloader_upgrade_cdn_cloudflare()
and filter_mathjaxloader_upgrade_mathjaxconfig_equal()
- get_assignments_with_rescaled_null_grades()
These have been kept because continue being used by restore:
- \core\task\refresh_mod_calendar_events_task
1. Fix sortable key for the qname/idnumber/tags column
2. Fix the default sort
3. Fix Behat test for sorting by idnumber. Added ID numbers for both
question A and question C in order for us to get consistent sorting
across DBs.
There is a proposed change MDL-66816 which will change the question
bank UI. This will break any Behat tests which uses low-level steps
like
When I click on "Duplicate" "link" in the "Test question" "table_row"
to perform an action on a question in the question bank. This commit
introduces a new step:
When I choose "Duplicate" action for "Test question" in the question bank
This commit also converts all core Behat tests to use the new step.
Question exports can be quite big, just like imports can. Therefore, we
should raise memory limits in the same way.
(Of course, eventually we should change the API, so that it writes one
questoin at a time, rather than building the whole export file contents
in RAM first, but that would be a much bigger change, and we need a
quick fix now.)
