1178829 - [META] Fix remaining idb web-platform-test failures

Status: NEW

(Core :: Storage: IndexedDB, defect, P3)

Description

[Boris Zbarsky [:bzbarsky]]

I just went through the tests that we are marked as failing at http://w3c.github.io/test-results/IndexedDB/all.html and a bunch were fixed by bent.  I fixed a few more in bug 1178803 and bug 1178806

That leaves the following test failures, all of which seem to be in the "test expects us to throw and we don't" bucket:

http://w3c-test.org/IndexedDB/idbcursor_advance_index9.htm
http://w3c-test.org/IndexedDB/idbcursor_advance_objectstore5.htm
http://w3c-test.org/IndexedDB/idbcursor_continue_index8.htm
http://w3c-test.org/IndexedDB/idbcursor_continue_objectstore6.htm
http://w3c-test.org/IndexedDB/idbcursor_delete_index5.htm
http://w3c-test.org/IndexedDB/idbcursor_delete_objectstore5.htm
http://w3c-test.org/IndexedDB/idbfactory_open9.htm (6 failures here) 

and these failures to do with ordering:

http://w3c-test.org/IndexedDB/idbobjectstore_createIndex6-event_order.htm
http://w3c-test.org/IndexedDB/idbobjectstore_createIndex7-event_order.htm 

that I decided I wasn't up to messing with today.

Comment 1

[Ben Turner (not reading bugmail, use the needinfo flag!)]

> http://w3c-test.org/IndexedDB/idbcursor_advance_index9.htm
> http://w3c-test.org/IndexedDB/idbcursor_advance_objectstore5.htm
> http://w3c-test.org/IndexedDB/idbcursor_continue_index8.htm
> http://w3c-test.org/IndexedDB/idbcursor_continue_objectstore6.htm

These are now bug 1179025.

Comment 2

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

We're down to

idbfactory_open10.htm
idbfactory_open9.htm.ini
idbobjectstore_createIndex6-event_order.htm
idbobjectstore_createIndex7-event_order.htm 
idbtransaction_objectStoreNames.html
interfaces.worker.js

Comment 3

[Bevis Tseng[:bevis][:btseng](Exited)]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=92fb681433bd

Comment 4

[Bevis Tseng[:bevis][:btseng](Exited)]

Summary of the remaining test cases:
Bug 1274938 depends on the update of IDB spec:
 - idbfactory_open10.htm
 - idbtransaction_objectStoreNames.html
Bug 1275496 depends on the plan of deprecating the StorageType in IDB:
 - idbfactory_open9.htm

Comment 5

[Bevis Tseng[:bevis][:btseng](Exited)]

Note: More failures are identified on wpt-dashboard: (All green in Chrome)
http://wpt.fyi/IndexedDB

Comment 6

[Bevis Tseng[:bevis][:btseng](Exited)]

Summary of the problems in our implementation according to the remaining wpt failures:
1. Clone before keypath evaluation:
(In current implementation,
 keypath evaluation is done on the passed-in JS object instead of the cloned one.
 See step 10 for the reason of evaluating on the cloned object in http://w3c.github.io/IndexedDB/#dom-idbobjectstore-put)
- clone-before-keypath-eval.html
- keypath-exceptions.htm
- idb-binary-key-detached.htm

2. Microtasks & promises (Bug 1193394):
(In step 4 of https://html.spec.whatwg.org/multipage/webappapis.html#perform-a-microtask-checkpoint,
 IndexedDB transactions has to be cleaned up at the end of microtask checkpoint.
 For now, we check this at the meta-stable state which is not part of the standard and is expected to be changed in bug 1193394.)
- event-dispatch-active-flag.html
- transaction-deactivation-timing.html
- upgrade-transaction-deactivation-timing.html
- upgrade-transaction-lifecycle-user-aborted.html

3. Index key extraction problem if compounded with |autoIncrement| primary key:
(Index key is extracted in the content processes but the auto-incremented primary key is generated later in the parent side.
 The object can be stored in the IDBObjectStore but is ignored in the IDBIndex:
 http://searchfox.org/mozilla-central/rev/20e41d4a61a8f5e34c9cf357304b78b3e9bced8a/dom/indexedDB/IDBObjectStore.cpp#1325-1328 )
- idbobjectstore_createIndex15-autoincrement.htm

4. Deprecate StorageType in IDBFactory.open (Bug 1276576)
- idbfactory_open9.htm

5. Intermintent failure (mostly on win8-opt build):
- fire-error-event-exception.html
- idbcursor-update-exception-order.htm
- idbdatabase-createObjectStore-exception-order.htm
- idbdatabase-deleteObjectStore-exception-order.htm
- idbobjectstore-deleteIndex-exception-order.html
- idbindex_getAll.html
- idbindex_getAllKeys.html
- idbobjectstore_getAll.html

Comment 7

[Marion Daly [:mdaly]]

Moving to DWS_NEXT

Comment 8

[Jens Stutte [:jstutte]]

:sgiesecke, most dependencies are resolved, maybe we just want to decide, if the remaining few are important enough to keep this bug?

Comment 9

[Simon Giesecke [:sg] [he/him]]

I am not sure what the original purpose of this meta-bug was. The use of "remaining" sounds as if this is meant to fix the failures based on a snapshot state. This has probably largely been addressed, and the meta-bug in this sense could be closed. On the other hand, it might make sense to have a permanent meta-bug that tracks all individual IndexedDB web-platform-test failure bugs, as they come and go.

Comment 10

[Jens Stutte [:jstutte]]

I'd just like to have clarity about the purpose of this bug. Should this then be:

1. a collection of all things that are gaps wrt the spec?
2. a collection of defects of our current implementation?
3. a mixture of both?

For 1) I would prefer to look at WPT dashboard and avoid manual tracking here, which will tend to be outdated anyway (btw, the bugs to fill such gaps should probably be enhancements, not defects). Once we decide to work on a specific set of gaps, we should then have an enclosing meta-bug with finite scope to track that set. So this bug can probably be closed.
For 2) we have the component, unless we have a lower granularity of commonalities of a smaller sub-set of defects. But I would expect all defects on implemented functionality to have the potential to harm our spec compliance. So this bug can probably be closed or should be made much more specific.
3) I deem neither manageable nor useful. So if we are not able to decide for 1) or 2), this bug can probably be closed, too.

:asuth, can you decide this (and in case adjust)? Thank you!

Comment 11

[Jens Stutte [:jstutte]]

:janv, can we ?

Comment 12

[Jan Varga [:janv]]

I don't think this should track all defects in our implementation. I think the original idea/goal was to fix all remaining failures in WPT tests.
WPT dashboard is nice, but a meta bug makes it easier to find bugs for specific issues.

Anyway, as Simon said, "Fix remaining idb web-platform-test failures" sounds more like a bug for specific snapshot.
So either we file a new bug or rename it to something more generic, like:
"IndexedDB web-platform-test failures tracking bug"

Comment 13

[Takanori MATSUURA]

All depending bugs have been fixed.