See attached. I am sure this is the source of quite a few mysterious fails. When debugging my update bug sometimes, rarely, it passes, but mostly it failes. I am sure it's sometimes the other way 'round.

Files: 
CommentFileSizeAuthor
order.txt1.77 KBchx

Comments

We have DependencyHookInvocationTest.php to test the order is correct, is it possible that we have a circular dependency?

Modules implementing hook_update_dependencies().

core/modules/block/block.install
core/modules/contact/contact.install
core/modules/overlay/overlay.install

All of these are concerned with getting the users_data sorted so there is no reason why user.module updates should come before the system.module ones

DependencyHookInvocationTest only tests when dependencies are enabled. The order does seem to be the same all the time but when it looks like I won't trust it...

Title:Random test failures: Updates are run in no particular orderUpdates are run in no particular order
Priority:Critical» Normal
Issue tags:+Random test failure

As discussed with @chx, changing the priority of this to normal.

I haven't seen random upgrade path test failures like we used to have frequently in quite a while and while this does look strange, it is IMHO neither critical nor major until we have proof that this actually causes problems with the upgrade path.