1890484 - When suffering HTTP 403/404, firefox/fenix sometimes show blank webpage, not error page

Status: RESOLVED DUPLICATE of bug 1945855

(Core :: Networking: HTTP, defect, P3)

Description

[Tom25519]

Steps to reproduce:

Note: this bug appears on Firefox for Windows 124.0.2 and Firefox for Android 124.2.0.

1. Open https://mp.weixin.qq.com/a/~XjoYJzjc0dPx85VleL2CVw~~?search_click_id=&version=28002f51&ascene=65&lang=zh_CN&pass_ticket=XU5t1bk%2FfSwJgirzvkyNi4byZT6QBFseQxXXBci6MH2c1InRtFlSQsMaty4L%2B4RKNfiIzMqIU4iiWxEzmEfbVA%3D%3D&wx_header=3

Actual results:

Blank webpage.

Expected results:

It should show "404 Not found", like Edge.

Comment 1

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

The Bugbug bot thinks this bug should belong to the 'Fenix::General' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[jackyzy823]

Attachment: ChromeForAndroid.png

The url return 404 with empty content. (Content-length is 0)

Chrome for Android shows a customized error page for this situation.

Comment 3

[Vasilica Tamas QA [:vtamas]]

Attachment: fenixblankpage.jpg

Thanks for reporting this issue!

I was able to reproduce it on Firefox Nightly 126 (2024-02-11), Firefox 125 Beta 9, Firefox 124.2.0 using a Samsung Galaxy A32 with Android 13. Confirming as New.

Comment 4

[Tom25519]

OS: Unspecified → Android
This bug also happens on desktop version firefox.

Comment 5

[Tom25519]

This bug also happens on https://lvlin.baidu.com/question/1743595071203483867.html (HTTP error 403).

Comment 6

[Sean Kim]

It seems like the content-length header is 1 instead of 0 (https://lvlin.baidu.com/question/1743595071203483867.html), otherwise it should invoke an error page, as Bug 1325876 is addressed.

Regarding the first URL (https://mp.weixin.qq.com/a/~XjoYJzjc0dPx85VleL2CVw~~?search_click_id=&version=28002f51&ascene=65&lang=zh_CN&pass_ticket=XU5t1bk%2FfSwJgirzvkyNi4byZT6QBFseQxXXBci6MH2c1InRtFlSQsMaty4L%2B4RKNfiIzMqIU4iiWxEzmEfbVA%3D%3D&wx_header=3), it should be addressed with Bug 1945855.

Comment 7

[Sean Kim]

Hello Randell,

Should we mark this bug as the duplicate of Bug 1945855, or vice versa?

Thank you

Comment 8

[Daniel Veditz [:dveditz]]

bug 1945855 describes what's happening with the link in comment 0 (no content-length: header), but the one in comment 5 is a little different (as you noted). If you duped this to bug 1945855 and then fixed it by only checking for an empty body when there's no content-length then that wouldn't be great. If you fix it by checking "there is no body, regardless of what the content-length does or doesn't say" then they could be the same.

The other bug is much easier to read, but is currently very specifically too narrow. And it doesn't have reproducing test links like this one.

A tangent that is separate from both of these bugs: Do we need to worry about error responses that DO have a body, but are "empty-ish"? Like, if the baidu link really returned a 1 byte response of a space or control character? It's may be rare in practice, but worth checking what Chrome does. If Chrome shows a blank page then great -- it's rare and we can stick with the "empty body" rule. If Chrome has some kind of "empty-ish" heuristic we might need to do the same. I hope they don't -- seems really edge-casey, but if they've done it then it must be common enough to confuse people.

Comment 9

[Daniel Veditz [:dveditz]]

bug 1945855 has been updated to make it clear it should handle both cases