808532 - File not found error for HTTP URL (2), can be fixed by a forced reload

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[Honza Bambas (:mayhemer)]

+++ This bug was initially created as a clone of Bug #771832 +++

I have to confirm this bug still exists.  I'm on Aurora.

I was submitting a new post to my blog, going initially through the main page.  After I submitted the post, I closed the page.  Then I realized I forgot to log off, and so I wanted to open the main page again, but it faild with File Not Found.

I was not clearing the cache, however, I was able to visit the same page few minutes ago.

Now I have problems with more then 2 other URLs that I was able to visit just a minute back.

Fortunately I have a 900MB backlog with append,timestamp,nsHttp:5,cache:5,nsOfflineCacheUpdate:5,nsHostResolver:5 I can privately submit for study.

Anybody interested?


Going to wipe the cache and see if that helps.

Comment 1

[Michal Novotny [:michal]]

(In reply to Honza Bambas (:mayhemer) from comment #0)
> Fortunately I have a 900MB backlog with
> append,timestamp,nsHttp:5,cache:5,nsOfflineCacheUpdate:5,nsHostResolver:5 I
> can privately submit for study.
> 
> Anybody interested?

I'd like to have a look at the log.

Comment 2

[Radek 'sysKin' Czyz]

I get it most frequently when going back/forward.
Chasing the cache entry in about:cache, it points me to a discrete file on disk. If I navigate there, the file is zero-sized.

This part looks the same as what bug 771832 fixed, except I'm definitely still getting it after the fix.

Comment 3

[Honza Bambas (:mayhemer)]

(In reply to Michal Novotny (:michal) from comment #1)
> I'd like to have a look at the log.

Sent privately.

Comment 4

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

I can sometimes also reproduce it. It happens very rarely, even on fresh install.

Comment 5

[Sid]

Reproducible on the latest trunk build. It happens several times a day, randomly, with no apparent cause, and only Ctrl+F5 helps.

Comment 6

[Katelyn Gadd (:kael)]

Copying and pasting from the other bug since apparently it was the wrong one (???):

Sadly I don't have reliable repro steps but I'm seeing the stylesheets for http://xboxforums.create.msdn.com/ fail to load around 50% of the time even though the web console's net view shows 304s for all the relevant assets. All I have to do is browse around through the various subforums.

One asset that frequently fails to load in this case is http://xboxforums.create.msdn.com/assets/css/main_Default_UpLevel.css?version=3.0.10335 and the cache failure also affects view-source, showing an error for 
Firefox can't find the file at view-source:http://xboxforums.create.msdn.com/assets/css/main_Default_UpLevel.css?version=3.0.10335 . Hopefully this is easy to reproduce for others as well.

Comment 7

[Alice0775 White]

I filed a new Bug 810781 about Comment 6

Comment 9

[Alfred Kayser]

May be there still need to be more cases where the file that was written too, needs to closed. mFD for output streams is only closed when Flush() is called, but it seems that in some error cases, Flush() is not called on close or destroy of the nsDiskCacheStream.

Comment 11

[bhavana bajaj [:bajaj]]

(In reply to Michal Novotny (:michal) from comment #1)
> (In reply to Honza Bambas (:mayhemer) from comment #0)
> > Fortunately I have a 900MB backlog with
> > append,timestamp,nsHttp:5,cache:5,nsOfflineCacheUpdate:5,nsHostResolver:5 I
> > can privately submit for study.
> > 
> > Anybody interested?
> 
> I'd like to have a look at the log.

Did anything from the log sent privately(as per comment #3) help out here ?

Comment 12

[Michal Novotny [:michal]]

Unfortunately, the log didn't provide enough information to find out what's the problem. I could find URLs with problematic entries and the scenario seemed to be identical to bug #771832. We don't log mFD manipulation enough in the cache code, so I wasn't able to find exact scenario. And I was unable to reproduce the bug locally with URLs that I found in the log.

Anyway, this particular code has changed significantly in bug #814010, so it might be that bug #814010 fixed this bug too.

Comment 13

[Frank Wein [:mcsmurf]]

At least I have not seen this bug happen in a long time now.

Comment 14

[Michael]

Same, this has been fixed for some time for me as well. It used to happen many, many times a day.

Comment 15

[Honza Bambas (:mayhemer)]

This has apparently been fixed by backout of bug 405407.