Re: [IndexedDB] Seeking status and plans

On Mon, Oct 20, 2014 at 9:32 AM, Arthur Barstow <art.barstow@gmail.com>
wrote:

> On 10/20/14 11:29 AM, Xiaoqian(Cindy) Wu wrote:
>
>> On 2014-10-15, at 5:10, Zhang, Zhiqiang <zhiqiang.zhang@intel.com
>> <mailto:zhiqiang.zhang@intel.com>> wrote:
>>
>>> See
>>> http://w3c.github.io/test-results/IndexedDB/all.html
>>> http://w3c.github.io/test-results/IndexedDB/complete-fails.html
>>> http://w3c.github.io/test-results/IndexedDB/less-than-2.html
>>>
>>> Are there anyone able to help run the tests with "additional browsers”?
>>>
>>
>> Here’s the report with IE11 and Safari IOS8 added. Hope it helps.
>>
>> See
>>
>> Pull Request: https://github.com/w3c/test-results/pull/8
>>
>> Updated Report:
>> http://siusin.github.io/test-results/IndexedDB/all.html
>> http://siusin.github.io/test-results/IndexedDB/less-than-2.html
>> http://siusin.github.io/test-results/IndexedDB/complete-fails.html
>>
>
> Thanks Zhiqiang and Xiaoqian!
>

Yes - a big thanks!

FYI, the majority of the "redness" in the "less-than-2" tests can be
attributed to a couple of reasons that don't significantly affect
compatibility:

   - interfaces.html - this idlharness.js-based test probes things like
   "does an attribute exist on the prototype of an interface object?" which
   Firefox gets right per WebIDL but the other browsers don't. This affects
   every API, not specific to Indexed DB. Chrome is attempting to fix without
   regressing performance. (I think there's more than one such issue revealed
   in the tests, and there may be actual implementation bugs too, will double
   check)
   - key_invalid.html - this assumes a particular prioritization of invalid
   input checks, and relates to
   https://www.w3.org/Bugs/Public/show_bug.cgi?id=26492

Accounting for those, the "less-than-2" tests show us to be in pretty good
shape across implementations.



>
> Sorry for my ignorance on the implementations, but would someone please
> confirm whether the IDB implementations for Ch40 and SF08 are the same or
> independent? (I assume they must be different or you would not have the
> SF08 data but I want to make sure.)
>

"It's Complicated."

Prior to the WebKit/Blink split, the bulk of Chromium's IDB implementation
existed in WebKit, although Safari did not have a functional implementation
(missing some glue). After the split, Chromium's code was significantly
refactored, but not rewritten. Safari implemented a completely new backing
store, but (from what I can tell) retained much of the front-end code and
at least some of the operation logic. We're definitely not able to share
bug fixes at this point due to code divergence.

So: we should have different results listed (since we're not
bug-compatible), but strictly speaking we might fail an "independent
implementations" criteria a standards body may be looking for.


>
> Simon - my recollection is Opera created an independent IDB implementation
> (before using Chromium). Is that true? If so, should we include that
> implementation in our testing?
>
> -Thanks, ArtB
>
>
>
>
>