Closed
Bug 1071290
Opened 10 years ago
Closed 10 years ago
Blob as one of the types in a Union crashes in workers
Categories
(Core :: DOM: Core & HTML, defect)
Tracking
()
RESOLVED
FIXED
mozilla36
People
(Reporter: nsm, Assigned: nsm)
References
Details
Attachments
(1 file)
13.24 KB,
patch
|
baku
:
review+
|
Details | Diff | Splinter Review |
The fetch spec's Request's body can be a blob [1].
Our codegen generates unwrapping code for main thread, but Blob has special handling on the worker to get a Blob from JS::Value, leading to crashes when a Blob is passed as the body.
[1]: https://fetch.spec.whatwg.org/#bodyinit
Comment 1•10 years ago
|
||
See bug 809899.
Sadly, it seems we'll compile stuff now, just incorrect stuff. I wonder whether we can do better here...
For the case of Blob, the right answer is to land bug 1047483, right?
Depends on: 1047483
Comment 2•10 years ago
|
||
Maybe what we can do is have union ctors assert mainthread if the union contains any interface type which has different worker and non-worker descriptors? That would at least catch the issue early in debug builds...
Assignee | ||
Updated•10 years ago
|
Blocks: dom-fetch-api
Assignee | ||
Comment 3•10 years ago
|
||
Just noting that this affects the Response object as well.
Comment 4•10 years ago
|
||
I think this bug is fixed by bug 1047483. Can you try again?
Flags: needinfo?(nsm.nikhil)
Assignee | ||
Comment 5•10 years ago
|
||
Thanks! That works.
Attachment #8502753 -
Flags: review?(amarchesini)
Assignee | ||
Updated•10 years ago
|
Assignee: nobody → nsm.nikhil
Status: NEW → ASSIGNED
Assignee | ||
Updated•10 years ago
|
Flags: needinfo?(nsm.nikhil)
Comment 6•10 years ago
|
||
Comment on attachment 8502753 [details] [diff] [review]
Allow Blobs in Fetch BodyInit
Review of attachment 8502753 [details] [diff] [review]:
-----------------------------------------------------------------
::: dom/fetch/Fetch.cpp
@@ +114,5 @@
> }
> }
>
> nsresult
> +ExtractByteStreamFromBody(const OwningArrayBufferOrArrayBufferViewOrBlobOrScalarValueStringOrURLSearchParams& aBodyInit,
wow. This is a really cool type name.
@@ +122,5 @@
> MOZ_ASSERT(aStream);
>
> if (aBodyInit.IsArrayBuffer()) {
> const ArrayBuffer& buf = aBodyInit.GetAsArrayBuffer();
> return ExtractFromArrayBuffer(buf, aStream);
So actually, just because you are returning value, I think we should have this:
if (aBodyInit.IsArrayBuffer()) {
...
return ...
}
if ( ... ) {
return ...
}
@@ +157,5 @@
> const ArrayBufferView& buf = aBodyInit.GetAsArrayBufferView();
> return ExtractFromArrayBufferView(buf, aStream);
> + } else if (aBodyInit.IsBlob()) {
> + const File& blob = aBodyInit.GetAsBlob();
> + return ExtractFromBlob(blob, aStream, aContentType);
save comment here.
Attachment #8502753 -
Flags: review?(amarchesini) → review+
Assignee | ||
Comment 7•10 years ago
|
||
I actually can't land this until the Response + Split InternalHeaders patch is reviewed :)
Assignee | ||
Comment 8•10 years ago
|
||
remote: https://hg.mozilla.org/integration/mozilla-inbound/rev/caa56adec4f5
Andrea,
I forgot to make the else-if to if change you suggested :/ Will there be performance savings in converting? If so, i can push a followup
Flags: needinfo?(amarchesini)
Comment 9•10 years ago
|
||
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla36
Comment 10•10 years ago
|
||
> I forgot to make the else-if to if change you suggested :/ Will there be
> performance savings in converting? If so, i can push a followup
This code is changing quickly. Let's do it next time we touch this code.
Flags: needinfo?(amarchesini)
Updated•6 years ago
|
Component: DOM → DOM: Core & HTML
You need to log in
before you can comment on or make changes to this bug.
Description
•