* nicer no-results search message

* also update SRP

* complicated prefix trim to solve 15247

* found regression, hah

* lint

* test

* Custom expression editor: fix Enter not commmitting the change

* Remove superfluous onChange

* Merge onDone and onUpdateAndCommit

* initial prototype of 'Explore results'

* use ViewButton

* use same button type as ViewFooter

* check for database

* extract explore results url into fn

The width should be fixed only when the editor is being used in the
notebook, and not from the sidebar.

There is this famous Firefox issue with empty contentEditable:
https://bugzilla.mozilla.org/show_bug.cgi?id=1020973

The workaround is to inject <br> when there is no expression (it will be
wiped out anyway on subsequent typing by the user).

* use schema list loader to check for number of schemas when displaying table context

* set loadingAndErrorWrapper to false to prevent UI flash

* fix up translation strings for metric/segments

* remove duplicate length check

* Record login history [WIP]

* WIP

* Add API endpoint and ugly React component

* Fix fn name

* Lookin good

* Test fixes 

* Prettier

* Test fix

* Test fixes 

* Better login history info

* Less ugly login history & tests

* Update migration numbers

* Deduplicate https-request? util function from metabase.server.middleware .session & .util

* Refactor -- new Ring request utils namespace

* Test fixes 

* Reorganize stuff a bit; API endpoint tests

* Tests for login/logout recording history

* Test fix 

* Test fixes 

* Test fix 

* Copy FE test fixture DB so it stops getting changed every time we run tests

* Test fixes 



* layout and code style tweaks

* Fix ns sorting

Co-authored-by: Kyle Doherty <5248953+kdoh@users.noreply.github.com>

* Fix typo in semantic migration

* Update semantic migrations for mysql compatibility

mysql updates as it goes, not in a row level transaction. So

```sql
update a = null,
       b = (case a ...)
where a is not null
```

mysql will use the new null value of a rather than the pre-udpate
value of a whereas pg and h2 will do row level transaction so you are
considering the pre-updated value of a.

Solution is to update these migrations to leave semantic type alone
and then one final set semantic_type = null where coercion_strategy is
not null. This is sensical since these two properties used to both be
in semantic_type, so if we extracted a coercion it could not have had
a meaningful semantic_type before.

* Look at effective-type in the UI

still want to fall back to base_type if effective_type is not
present. Our inference doesn't set effective_type at the moment as
this change was super invasive and led to conflicts to other refactors
at the time. In the future almost nothing except for query processors
should care about base_type and only look at the (required)
effective_type.

* Check base and semantic (not base and effective) when no effective_type

* Correct typo in isNumericBaseType

was incorrectly checking effective_type instead of either effective or base

It does not make sense anyway for the web browser to suggest corrections
for function names etc.

* add throttle to ExplicitSize height/width recalc

* add comment

* Prioritize exact matches when querying the DB for a search

* Limit length of search input

* Add some tests

* Add card.description to search results

Merge x.38.2 -> master

* Add `cypress-skip-test` library

This is the official Cypress library for conditionally skipping/running tests.
https://github.com/cypress-io/cypress-skip-test

* Enable `cy.onlyOn` and `cy.skipOn` custom commands in all tests

Reference: https://github.com/cypress-io/cypress-skip-test#example

Merge pull request from GHSA-m8w4-6wxv-8v6m

* Fix sandboxed Field Values regression

* Test fixes 

* Dox fix 

* Port code for building i18n resources to Clojure build scripts; remove dep on gettext

* Use Metabase classloader

* Test fixes 



* Bump adoptopenjdk version

* Delete gettext dependency from Dockerfile

* Update developers-guide.md

* Address PR feedback

* Fix frontend singular msgstr format

Co-authored-by: Luis Paolini <paoliniluis@gmail.com>

Merge release x.38.x -> master 2021-03-16

Merge pull request from GHSA-pxfh-93j9-h745

Check read permission on the instance from get

Check write permission on the instance from revert

Rewrite expectations to clojure.test in revision_test.clj

Remove some ugliness (in the delay) and switch to fixtures instead

Add new test in revision_test.clj for preventing unauthorized reversions

* Convert even more old-style tests to the new style

* Test fixes 

* Search dashboard descriptions

* Restrict dashboardcard_count score to cards

* Restrict pinned score to appropriate models

* Add fullness-scorer

* Kill a silly function

* D'oh

* Get rid of text-score-max 

* Change webpack config

Use the parallel:true flag in the Uglify plugin to use all cores in a CPU for minification which greatly improves front end build time. You will now be able to take advantage of that watercooler you asked Santa for Christmas

* Kick the build

Co-authored-by: Tim Macdonald <tsmacdonald@gmail.com>

Merge 38.1 -> master part 5

# Conflicts:
#	enterprise/frontend/test/metabase-enterprise/audit/auditing.cy.spec.js
#	enterprise/frontend/test/metabase-enterprise/sandboxes/sandboxes.cy.spec.js
#	frontend/test/metabase/scenarios/admin/settings/localization.cy.spec.js
#	frontend/test/metabase/scenarios/dashboard/chained-filters.cy.spec.js
#	frontend/test/metabase/scenarios/dashboard/dashboard.cy.spec.js
#	frontend/test/metabase/scenarios/public/public.cy.spec.js
#	frontend/test/metabase/scenarios/question/filter.cy.spec.js
#	frontend/test/metabase/scenarios/question/native.cy.spec.js
#	frontend/test/metabase/scenarios/question/view.cy.spec.js

Provide means to add entities to the application db using specmonstah library.

There are 4 main parts to it:
- specs for column types (id, names, emails...). Used to generate random values.
- specs for entities. Aggregations of types
- schema. Structure of the dependency graph. Used by specmonstah to create the
  entities in the right order, and to link fks between them.
- insert. 
    - Generating the entity graph, then removing the `id`s (so
      the underlying db adds them and the id's don't clash with already existing ones).
    - Adjust field values that have constraints besides DB ones.
    - insert to the db using `toucan`.