This idea has floated around for quite some time. The testbot currently depends on a specific form layout in order to install Drupal. When changes are made to the installer the testbot must also be changed. I would like to separate the two pieces so the installer can be changed without effecting the testbot.
I propose we add a script for installing Drupal with a set of assumptions made about the environment provided. In other words, assume we have a database named "drupal" with credentials "drupal" and "drupal", etc.
Not sure if their is any reasons to generate a random password of to use the same like "admin" and "drupal" or "drupal and "drupal".
Dries mentioned that we should be sure this doesn't end up as security risk in that people can install Drupal without some sort of credentials. I think if we provide a script only runnable form command line that should be all the credentials we need.
Also Dries agreed that we could make it a drush script (or any power-user feature), so I will shoot for that. Otherwise this should be quite simple and allow the installer to be changed and tested without waiting for the testbot.
| Comment | File | Size | Author |
|---|---|---|---|
| #3 | 1317528-testing-install.patch | 2.16 KB | boombatower |
| #2 | 1317528-testing-install.patch | 2.19 KB | boombatower |
Comments
Comment #2
boombatower commentedNeed #1317562: Include core modules in command search path and #673020: Tests for adding the Content-Language HTTP header to the generated page (core bug).
Had a long chat chx, mosh, and samkottler in BADCamp coder lounge about drush commands in core. Seems like everyone is happy.
Comment #3
boombatower commentedUpdate to remove DRUPAL_ROOT which is left over from before drush script.
Comment #4
boombatower commentedLooks like moshe has fixed the drush patch, so we just need the bug breaking CLI installation in Drupal 8 fixed.
Comment #5
boombatower commented#1317578: Take advantage of core `drush simpletest-install`
Comment #6
sunThis adds another dependency on drush for running tests, which is a very complex system on its own.
I wonder, couldn't we achieve the same result without relying on drush?
Comment #7
boombatower commentedSure, just figured since we all agreed we could use drush for command we would start with this and convert them if/when people got bored. This doesn't do anything drush specific.
Comment #8
moshe weitzman commentedSimpletest now uses Drupal's non-interactive installer - install_drupal()
Comment #9
David_Rothstein commentedBut the testbot doesn't: http://drupalcode.org/project/project_issue_file_review.git/blob/875b3fd...
As I understand it, the script being proposed here would be for the testbot to use.
Comment #10
moshe weitzman commentedRight you are. Apparently, pift is using Drush site-install now to setup the parent Drupal site. This has become a major problem with D8 kernel refactoring so perhaps this issue should be revived. See #2016629-13: Refactor bootstrap to better utilize the kernel
Comment #11
rfayPIFR 6.x-3.x, which I expect to get rolled out to all testbots in the short term, is in fact now using drush si to install. The point-and-click install using D6 simpletest just was too unstable and wasn't maintainable. So with this there's no testing of the manual web-based install process.
Comment #12
sunYes, this (circular) dependency causes major problems:
#2206501: Remove dependency on Drush from test reviews
FWIW, we have real/proper tests for the interactive installer in D8 core now, so this legacy convenience feature of PIFR is no longer necessary (for D8).