Closed Bug 757284 Opened 12 years ago Closed 12 years ago

Issues with (large?) uploads stopping halfway (youtube, mediafire)

Categories

(Core :: DOM: Core & HTML, defect)

13 Branch
x86
macOS
defect
Not set
normal

Tracking

()

RESOLVED FIXED
Tracking Status
firefox13 + verified
firefox14 + verified
firefox15 + verified

People

(Reporter: cww, Assigned: khuey)

References

Details

(Keywords: regression, reproducible, Whiteboard: [qa+])

Attachments

(2 files)

From input:


It looks like uploads aren't working or stopping part way. Seems to be larger files(?) and 13 only.  Possibly not reliably reproducible.

http://input.mozilla.org/opinion/2940833 << mediafire
http://input.mozilla.org/opinion/2940450 << mediafire  
http://input.mozilla.org/opinion/2938682 << shutterfly

And a lot of youtube:

http://input.mozilla.org/en-US/?product=firefox&sentiment=sad&date_end=&date_start=&version=13.0&q=upload+youtube

And some facebook:

http://input.mozilla.org/en-US/?product=firefox&sentiment=sad&date_end=&date_start=&version=13.0&q=upload+facebook

UNCO pending QA.
On Mac 10.7.x from two different locations, 300mb+ upload, on Youtube, Vimeo, Smugmug:

- Fx12: Works
- Chrome: Works
- Fx13b1-4: Stops at 33% on Youtube, reproducible all the time.
- Fx15 and it also stops at 33% on Youtube

Using a 100mb+ file on mediafire.com:

- Fx12: works
- Fx13b4: never initiates the upload, it just stays at "Queued"
Status: UNCONFIRMED → NEW
Ever confirmed: true
Keywords: regression
Component: General → Networking
Product: Firefox → Core
QA Contact: general → networking
Keywords: reproducible
yumm.. I'll take a look and try and repro to see what's happening tonight or tomorrow.
Assignee: nobody → mcmanus
interesting the youtube.com case is hijacking http 308 for a non-redirect purpose. I thought we heard that was over?

but its not the cause of this bug, as the new 308 handling isn't included until FF14 and I don't think it would have any impact without the location header anyhow

2082469632[7f0a8de27030]: http response [
2082469632[7f0a8de27030]:   HTTP/1.1 308 Resume Incomplete
2082469632[7f0a8de27030]:   Server: HTTP Upload Server Built on May 16 2012 12:03:24 (1337195004)
2082469632[7f0a8de27030]:   Content-Type: text/html; charset=utf-8
2082469632[7f0a8de27030]:   Access-Control-Allow-Origin: http://www.youtube.com
2082469632[7f0a8de27030]:   Access-Control-Expose-Headers: Range, X-Range-MD5, Location, X-GUploader-Stats-URL
2082469632[7f0a8de27030]:   Access-Control-Allow-Credentials: true
2082469632[7f0a8de27030]:   Range: bytes=0-104857599
2082469632[7f0a8de27030]:   X-Range-MD5: 9e2ebb06c114ba0f5ca24812a6ee83af
2082469632[7f0a8de27030]:   Date: Wed, 23 May 2012 01:40:08 GMT
2082469632[7f0a8de27030]:   Pragma: no-cache
2082469632[7f0a8de27030]:   Expires: Fri, 01 Jan 1990 00:00:00 GMT
2082469632[7f0a8de27030]:   Cache-Control: no-cache, no-store, must-revalidate
2082469632[7f0a8de27030]:   Content-Length: 720
2082469632[7f0a8de27030]: ]
Blocks: 714302
The change we *did* in FF 13 with respect to redirects was 598304 (which changed method rewriting).

Would be good to have traces of both succeeding and failing uploads.
(In reply to Julian Reschke from comment #4)
> The change we *did* in FF 13 with respect to redirects was 598304 (which
> changed method rewriting).
> 

Thanks.. I made a build with that reverted we still failed the mediafire test (that's the one I'm using to bisect as its faster).. so probly not that.
I've bisected this to be driven from bug 725289 - which changes nsIDOMFile::mozSlice to be nsIDOMFile::slice.. presumably the affected sites are using that API.

it repros with either mediafire or youtube, but testing is easier with mediafire because you can use any size document and the free personal account they offer. My youtube test is 163MB and it stalls around the 100MB mark.
Assignee: mcmanus → nobody
Blocks: 725289
Component: Networking → DOM
QA Contact: networking → general
Summary: Issues with (large?) uploads stopping halfway → Issues with (large?) uploads stopping halfway (youtube, mediafire)
No longer blocks: 714302
Assignee: nobody → khuey
Status: NEW → ASSIGNED
Attachment #626888 - Flags: review?(jonas)
Does comment 6 satisfy QAWANTED and REGRESSIONWINDOW-WANTED?
Juan, since you were able to reproduce this earlier, can you please check the Feb 16th and 18th Nightly builds? That would at least confirm comment 6. Thanks
This is one of our two most critical FF13 issues. Although code freeze is 5/25, we can land a confirmed fix on Monday prior to the go to build (if we want to get testing on m-c first).
(In reply to Anthony Hughes, Mozilla QA (irc: ashughes) from comment #10)
> Juan, since you were able to reproduce this earlier, can you please check
> the Feb 16th and 18th Nightly builds? That would at least confirm comment 6.
> Thanks

I tested with youtube.com:

20120216031231: Works
20120218031156: Does not work

However I also tried mediafire and both worked.
Thanks Juan. I'm thinking that at least confirms Patrick's suspicion of bug 725289 and qualifies qawanted. Alex et al, please re-add the keyword if there is more QA can test around this. I'm suspecting at this point we just need to wait for and test a fix though.
Comment on attachment 626990 [details] [diff] [review]
Patch for beta

[Triage Comment]
Pre-approving for Beta. We'll back this out if any regressions are found.
Attachment #626990 - Flags: approval-mozilla-beta+
https://hg.mozilla.org/mozilla-central/rev/7379306d1ed8
Status: ASSIGNED → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Comment on attachment 626990 [details] [diff] [review]
Patch for beta

We should get this in on aurora too after it's confirmed it works

[Approval Request Comment]
Bug caused by (feature/regressing bug #): bug 725289
User impact if declined: Uploads to various things may break
Testing completed (on m-c, etc.): TBD
Risk to taking this patch (and alternatives if risky): Low risk
String or UUID changes made by this patch: None (uses a separate interface)
Attachment #626990 - Flags: approval-mozilla-aurora?
Comment on attachment 626888 [details] [diff] [review]
Restore mozSlice, with a warning (for trunk)

Review of attachment 626888 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/base/src/nsDOMFile.cpp
@@ +250,5 @@
> +    nsCOMPtr<nsPIDOMWindow> window = do_QueryInterface(sgo);
> +    if (window) {
> +      nsCOMPtr<nsIDocument> document =
> +        do_QueryInterface(window->GetExtantDocument());
> +      if (document) {

window->GetExtantDoc() would have saved you a QI, fwiw.
Attachment #626990 - Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Why on earth do we have a GetExtantDoc and a GetExtantDocument?
GetExtantDoc has been added by bug 756381 very recently for optimization.
Whiteboard: [qa+]
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:13.0) Gecko/20100101 Firefox/13.0 (beta 6)

Verified the fix on above build: successfully uploded 900+ MB video on Youtube and Facebook (the video was removed after upload for being too long) and ~114MB video on Mediafire.

Marking verifioed for Firefox 13
[Triage Comment]
Reminder that the merge day is tomorrow afternoon PT, so please land this to Aurora (14) before then.
Actually, the patch failed to build so I had to back it out. Seemed like a #include issue.
looks like the flag got reset, setting it back to 'fixed'.
I've verified this on;
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:14.0) Gecko/20100101 Firefox/14.0 beta 6 (build2)

Upload was performed without stopping or lagging on Youtube and also facebook using a file that had more than 900MB.

Setting the flag to Verified on Firefox 14.
I've verified this on:
Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7; rv:15.0) Gecko/20100101 Firefox/15.0 beta 1

The upload on youtube and facebook was made without any issues.

Setting the flag to Verified.
Depends on: 961804
Component: DOM → DOM: Core & HTML
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: