1906323 - Fenix can silently fails to load links on network failure with no error notification (loss of connection, loss of WAN connectivity

Status: NEW

(Firefox for Android :: Browser Engine, defect)

Description

[Andrew Creskey [:acreskey]]

On a network with no WAN link,Fenix will silently fail to load links with no error notification.

For comparison, Chrome will generally display an error page, e.g. "This site can't be reached".

Steps to reproduce

1. Load a top news site (e.g. bbc.com, cnn.com, ) in Fenix
2. Disconnect your wireless access point from the WAN
3. Click on a link

Expected behavior

An error is displayed or other indication that the pageload failed
In Chrome, the error page displayed is generally "This site can't be reached"

Actual behavior

The pageload progress bar disappears and thus the pageload silently fails without notification

Here's an example profile with nsHttp:5,nsSocketTransport:5,profilermarkers logs as markers.
https://share.firefox.dev/3XL4yil

From the user's perspective, the page load silently fails after about 47 seconds.

Note that while these particular steps to reproduce are artificial (i.e. disconnecting the access point from the WAN), scenarios in which the access point loses connection to WAN are common for mobile users.
e.g.
• wifi on a public transit as it travels between areas of coverage,
• events/venues: device may be able to connect to venue's wifi network, but there may not be stable WAN uplink to handle all users

Device information

• Firefox version: Fx129, also seeing same behaviour in Fx127, release channel
• Android device model: Samsung A54
• Android OS version: 14

Comment 1

[Andrew Creskey [:acreskey]]

We likely will need to move this bug to the ideal product/component.

Comment 2

[BugBot [:suhaib / :marco/ :calixte]]

The severity field is not set for this bug.
:boek, could you have a look please?

For more information, please visit BugBot documentation.

Comment 3

[Andrew Creskey [:acreskey]]

Updated title because the silent failure can also occur on loss of connection to network.

In this example the connection to the access point was lost.
https://share.firefox.dev/4cyn9ll

nsHttp logs included as profiler markers.

Comment 4

[Andrew Creskey [:acreskey]]

Attachment: silent_pageload_fail.mp4

In this recording I clicked on the link for the main article.
We can see the pageload starts (progress bar at about 25%).
But as the network transitions, the pageload silently fails.

Comment 5

[Andrew Creskey [:acreskey]]

Attachment: chrome_change_of_network.mp4

For comparison, we can see how Chrome handles the same scenario:

The pageload is initiated, progress bar displayed
Chrome displays "No internet" page on loss of connection to initial interface
Chrome automatically resumes pageload on change of network
Page successfully completes load

Comment 6

[Andrew Creskey [:acreskey]]

A profile (with nsHttp and related logs) where we have the pageload silently fail
https://share.firefox.dev/3SXbVQw

And one where we do actually get the "No internet" error page
https://share.firefox.dev/3SU6uC0

The request is "https://www.nytimes.com/2024/08/21/business/economy/harris-price-gouging-ban-groceries.html"

Comment 7

[Andrew Creskey [:acreskey]]

If I have this right, the request fails here with this status:

nsHttpChannel::OnStopRequest [this=6bfdf74600 request=6bac822880 status=804b0002]

NS_BINDING_ABORTED = 0x804B0002

Comment 8

[Andrew Creskey [:acreskey]]

Verified that this bug is still happening in the latest Fenix nightly.
It looks like it's often taking 60 seconds for the link-click load to silently fail.
Chrome is displaying an error page within 6-30 seconds.

For the necko side, let's first verify that we are correctly returning the failure.