1085060 - XMLHttpRequest throws when sending a Uint8Array from a web worker

Status: RESOLVED FIXED

(Core :: DOM: Workers, defect, P3)

Description

[Christopher Lund]

Attachment: testcase - worker script

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_8_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/37.0.2062.120 Safari/537.36

Steps to reproduce:

1. Open the attached testcase or visit it online at http://plnkr.co/edit/vooLd13AygHzTrQcrvYr
2. Click "Test Main Thread" to perform 3 POSTs using an XMLHttpRequest, one each for a string, Uint8Array, and ArrayBuffer. See the results below the buttons in the "Main Thread Results" section.
3. Click "Test Worker Thread" to do the same using a web worker. See the results in the "Worker Thread Results" section.

In addition to the 36.0a1 (2014-10-18) nightly, this is also seen in 32.0.3 and was observed on both Mac and Windows (64-bit) installations. Chrome 37, Safari 6.0.5, and Internet Explorer 11 pass all 6 test cases.


Actual results:

On the main thread, the XMLHttpRequest accepts all three data types as shown by the output "ok". In the web worker, the string and ArrayBuffer pass with "ok" but the Uint8Array fails with the exception: "Data conversion error"  nsresult: "0x80460001 (NS_ERROR_CANNOT_CONVERT_DATA)".


Expected results:

As the main thread succeeds for all three data types, the web worker should have succeeded for the types as well.

Comment 1

[Christopher Lund]

Attachment: testcase - HTML page

Comment 2

[Christopher Lund]

Attachment: testcase - HTML Page

Comment 3

[Christopher Lund]

Attachment: testcase - HTML page

Comment 4

[Christopher Lund]

Attachment: testcase - HTML page

Comment 5

[Christopher Lund]

Attachment: testcase - HTML page

Comment 6

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

This is a problem in the variant handling.  Attempting to turn the variant into a string is failing for some reason.

Comment 7

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

XPCVariant doesn't handle TypedArrays.  It ends up creating an XPCWrappedJS for them.  It does have code to handle regular Arrays though.  Should this be something we can expect XPConnect to handle?

Comment 8

[Bobby Holley (:bholley)]

(In reply to Kyle Huey [:khuey] ([e-mail]) from comment #7)
> XPCVariant doesn't handle TypedArrays.  It ends up creating an XPCWrappedJS
> for them.  It does have code to handle regular Arrays though.  Should this
> be something we can expect XPConnect to handle?

We could add support yes, though I doubt that teaching XPConnect how to convert a TypedArray into an XPCOM array (which is what adding support in XPCVariant would mean) is actually what we want here. We just want to treat it as a jsval, which nsIVariant doesn't do.

Shouldn't SendRunnable::MainThreadRun create the appropriate RequestBody directly rather than using the nsIVariant stuff?

Comment 9

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

(In reply to Bobby Holley (:bholley) from comment #8)
> Shouldn't SendRunnable::MainThreadRun create the appropriate RequestBody
> directly rather than using the nsIVariant stuff?

Yeah.  I was asking about the XPConnect stuff mostly for completeness.

Comment 10

[Thomas Wisniewski]

This will be resolved by one of the patches pending to land in bug 792808.

Comment 12

[Boris Zbarsky [:bzbarsky]]

Attachment: Testcase updated to new Bugzilla attachment hostnames

Comment 13

[Boris Zbarsky [:bzbarsky]]

Marion, the cleanup in bug 792808 doesn't seem like it's really going to land anytime.  Should we really block this web compat issue on that cleanup?  Seems like we should reopen and fix this on its own terms, unless we're going to push bug 792808 over the line.

Comment 14

[Marion Daly [:mdaly]]

You raise a good point.

Comment 15

[Thomas Wisniewski [:twisniewski]]

bz, this is safe to close as a dupe of bug 792808 (now that it has landed), right?

Comment 16

[Boris Zbarsky [:bzbarsky]]

Assuming the testcases here are fixed and we have wpt coverage for them, yes.

Comment 17

[Thomas Wisniewski [:twisniewski]]

I just confirmed that the testcase here does work as expected now, and that there are relevant WPTs which can be easily refactored to also work with workers/sharedworkers. I'll request a review for that refactoring shortly.

Comment 18

[Thomas Wisniewski [:twisniewski]]

Attachment: Bug 1085060 - update WPTs for XHR sends to also test with workers; r?annevk

update WPTs for XHR sends to also test with workers

Comment 19

[Thomas Wisniewski [:twisniewski]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=fc4b8076b4d05bcbf204167d967a0bffb242fcc0

Comment 20

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/01cbcad8bb67
update WPTs for XHR sends to also test with workers; r=annevk

Comment 21

[Natalia Csoregi [:nataliaCs]]

https://hg.mozilla.org/mozilla-central/rev/01cbcad8bb67

Comment 22

[Web Platform Test Sync Bot [:wpt-sync] (Matrix: #interop:mozilla.org)]

Created web-platform-tests PR https://github.com/web-platform-tests/wpt/pull/15897 for changes under testing/web-platform/tests