At this point, I think dropping the bootstrap flag from system table would just work as the bootstrap invoker calls module_implements which is registry based and loads on demand. I would suspect that calling module_invoke_all would work as well, we can drop bootstrap_invoke_all as well. module_list can lose its $bootstrap flag as well as there is no sensible notion of that any more. Once that's done the function will return the list of enabled modules so let's rename it to that. We can have a module_list_loaded which just is a wrapper around drupal_load.... 'cos you know registry_check_code really should use drupal_load. if this awfully resembles the total module revamp, well, we are getting there :)
| Comment | File | Size | Author |
|---|---|---|---|
| #3 | bootstrap.begone.patch | 13.85 KB | Anonymous (not verified) |
| #2 | bootstrap.begone.patch | 13.1 KB | Anonymous (not verified) |
Comments
Comment #1
Anonymous (not verified) commentedi'll take this one on.
Comment #2
Anonymous (not verified) commentedhere's a first cut.
Comment #3
Anonymous (not verified) commentedand, here's a second go.
updated to keep up with HEAD, and based on some feedback from chx in #drupal - i'd missed a chunk to be removed from
module_rebuild_cache.Comment #4
Anonymous (not verified) commentedfrom #drupal:
so, setting to CNW.
Comment #5
Anonymous (not verified) commentedhmm, seems we already test that
hook_bootandhook_exitget called as they should inHookBootExitTestCase.setting back to code needs review, patch updated only to keep up with HEAD, all tests pass.
Comment #6
dries commentedThis looks sane. Nice clean-up. Committed to CVS HEAD. Thanks.