1003100 - pushState() in iframe fails when reloading with CTRL+F5

Status: RESOLVED FIXED

(Core :: DOM: Navigation, defect)

Description

[Christoph Sieghart]

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:28.0) Gecko/20100101 Firefox/28.0 (Beta/Release)
Build ID: 2014031500

Steps to reproduce:

1. Open console
2. Go to http://jsbin.com/vadoloda
3. See log message "iframe content"
4. Reload Page by pressing F5
5. No NS_ERROR_FAILURE occurs
6. Reload Page by pressing CTRL+F5
7. An NS_ERROR_FAILURE occurs



Actual results:

NS_ERROR_FAILURE:
window.history.pushState(null, null, "#foo")


Expected results:

No NS_ERROR_FAILURE should happen.

Comment 1

[Bogdan Maris, Desktop Test Engineering]

Reproduced on Windows 7 64bit and Mac OS X 10.9.2 using latest Nightly.

Comment 2

[Jonathon Munsell]

Still valid and 100% reproducible in latest release of FireFox (v37.0.1 at time of this writing). Only appears to a occur with a hard refresh(`shift + f5` on Windows, `cmd + shift + r` on Mac), cannot reproduce with standard refresh(e.g. click refresh button in browser, `f5` on Windows, `cmd + r` on Mac) or on first time load of page.

Comment 3

[CircleCode]

It may be interseting to note that it still occurs with firefox nightly (41 at the time of this writing).

In my case, I have an application that answers to user action by
1. creating an iframe
2. calling window.pushState on the new iframe

It works perfectly unless I have used ctrl + f5 before. Then, when I use my button (even  if I wait a long after the hard reload), I get an NS_ERROR_FAILURE

Comment 4

[Burak Yiğit Kaya [:BYK]]

We are getting this pretty frequently from our users too at Disqus.

Comment 5

[Matthew N. [:MattN]]

Attachment: Testcase to run from a file

The problem is that we're failing at:
NS_ENSURE_TRUE(mOSHE) failed: file /docshell/base/nsDocShell.cpp, line 11483

https://mxr.mozilla.org/mozilla-central/source/docshell/base/nsDocShell.cpp?rev=982d9a863989#11483

smaug, any chance you or someone else could take a quick look at this to at least see if there is a simple fix?

Comment 6

[Matthew N. [:MattN]]

Comment on attachment 8663788 [details]
Testcase to run from a file

For some reason this testcase doesn't work from BMO but works from a file URI. In case it isn't clear, this isn't specific to @srcdoc usage, I just used that to make the testcase self-contained.

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

That testcase with srcdoc doesn't seem to work, but let me try to find some time for this, perhaps next week.

Comment 8

[Olli Pettay [:smaug][[e-mail]]]

Attachment: v1

This is super fragile, but the issue is that we propagate loadtype to subframes, even though those are in shift-reload case newly created iframes and should get an mLSHE (which is then later moved to mOSHE).

https://treeherder.mozilla.org/#/jobs?repo=try&revision=85228661f4a3

The issue with srcdoc is https://bugzilla.mozilla.org/show_bug.cgi?id=1215298

Comment 9

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8674502 [details] [diff] [review]
v1

r=let's-pretend-I'd-have-noticed-if-there-were-something-wrong-here.

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

And based on try the assertion is too strong. We don't create a new entry anyway if we already have
mLSHE, so removing !mLSHE from the assertion.

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

Attachment: weaker assertion

https://treeherder.mozilla.org/#/jobs?repo=try&revision=93686ed30a12

Comment 12

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/4f4e54b55fc8

Comment 13

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/4f4e54b55fc8