777621 - Move test_write_tabs_503 into memcached-specific tests

Status: VERIFIED FIXED

(Cloud Services Graveyard :: Server: Sync, defect)

Description

[Ryan Kelly [:rfkelly]]

The "test_write_tabs_503" tests specific behaviour of the memcached backend, but it included in the base generic tests for syncstorage.  I should move it into the memcached-specific tests so that it gets skipped properly when memcache is not available.

Comment 1

[Ryan Kelly [:rfkelly]]

Attachment: patch moving the tab-related functional tests

Moved the relevant tests into TestMemcachedStorage class and cleaned them up a bit.

Comment 2

[Ryan Kelly [:rfkelly]]

Attachment: patch to fix locking of cache-only collections

Naturally, re-jiggering these tests has revealed a bug.

The memcached backend cannot lock cache-only collections by taking locks in the database, because those collections don't exist in the database.  Instead it must use the cache-based locking mechanism.

Comment 3

[Toby Elliott [:telliott]]

Comment on attachment 646403 [details] [diff] [review]
patch to fix locking of cache-only collections

Is there really not a generic way in python to do _as_list? Seems like that belongs straight in a util lib somewhere.

Comment 4

[Toby Elliott [:telliott]]

Comment on attachment 646401 [details] [diff] [review]
patch moving the tab-related functional tests

The diff made this a little hard to check, but I was unable to spot notable differences between the two. If it's just a move, then r+.

Comment 5

[Ryan Kelly [:rfkelly]]

(In reply to Toby Elliott [:telliott] from comment #4)
> Comment on attachment 646401 [details] [diff] [review]
> patch moving the tab-related functional tests
> 
> The diff made this a little hard to check, but I was unable to spot notable
> differences between the two. If it's just a move, then r+.

The "for key in self.config.registry" loops that setup/teardown the BadCache have changed slightly.  The new one is a lot simpler, because we now know that we've already got a MemcachedStorage instance, rather than having to create a new one.

Comment 6

[Ryan Kelly [:rfkelly]]

https://github.com/mozilla-services/server-syncstorage/commit/eeb18c4323442d3078cc990a9fdc0c2019dee0fa

Comment 7

[Ryan Kelly [:rfkelly]]

(In reply to Toby Elliott [:telliott] from comment #3)
> Is there really not a generic way in python to do _as_list? Seems like that
> belongs straight in a util lib somewhere.

Aha, I see there is a pyramid.settings.aslist function that does the same job.  I'll switch over to that.

Comment 8

[Ryan Kelly [:rfkelly]]

Attachment: patch to use aslist helper from pyramid

Comment 9

[James Bonacci [:jbonacci]]

Verified as fixed with the latest code (just pulled today).
Also noting that :telliott just "+"-ed the code review (good timing)

Comment 10

[Ryan Kelly [:rfkelly]]

aslist() helper change committed in https://github.com/mozilla-services/server-syncstorage/commit/78ff2f06e04a7fd8957a26dec9fb3680d058e78d