Bugzilla – Bug 828779
Firefox refresh badly on Bugzilla
Last modified: 2013-11-12 14:46:39 UTC
What I describe here is rare, but it has happened to me thrice. I write a comment on a Bugzilla, the Novell Bugzilla. Somebody answers, and changes the component, from 'A' to 'B'. I notice the email, go to the Firefox tab that still has that report open, and click the refresh circular arrow on FF. Firefox refreshes the page, but the component still shows 'A' (unnoticed by me). If I write a comment now, the component is changed to 'A', even if I did not change it myself, my page never displayed the correct component 'B'. Result: people blame me for undoing their changes, because it had already been decided that component should be 'B'. In fact, I have to press and hold the shift key while I click on refresh to force FF to really refresh the page, so that the component of the Bugzilla refreshes and displays 'B'. Reports where this has happened to me: https://bugzilla.novell.com/show_bug.cgi?id=827331#c5 https://bugzilla.novell.com/show_bug.cgi?id=827196#c10 https://bugzilla.novell.com/show_bug.cgi?id=825339#c24 Please find out if this is a Firefox problem or a Bugzilla problem and solve as appropriate. If you have a manner of testing this and I can help, please tell me. I'm getting tired of being accused of messing the fields in Bugzilla, and I don't know if they believe me or not. :-(
FYI: As far as I got told by a colleague it could be mainly a Bugzilla issue. Reason: When you only add a comment and click the [Commit] button in the Bugzilla web from in your browser, then not only your added comment is sent to the Bugzilla server but all data is sent to it. Therefore when you only add a comment and click the [Commit] button also the other fields (e.g. also the "Component" field) is sent to the Bugzilla server so that on the Bugzilla server not only your comment gets added but also all other fields are set to the values that you have sent. I think you should get a "mid air collision" notification in such cases so that you are at least informed but I don't know the exact conditions when one gets such a notification. I think if the Bugzilla server could remember the set of values that it had sent to you and compare it with the set of new values that it gets from you and only update those values that you actually changed, then it could "just work correctly" even without any manual refresh action in your browser. Some drawbacks: For how long should the Bugzilla server remember the set of values that it had sent to anyone who has update permissions? One hour? 12 hours? How much memory would the Bugzilla server need to do this? How much data processing overhead would this cause? How much would this slow down the Bugzilla server?
But you see, the extra problem I have, and I don't know if you do or is unique to me, is that when I click refresh to see your comment, the comments are refreshed, but the several fields: status, product, components, platform, etc, are not. THAT is the important difference. There can not be mid-air-collisions, because I did refresh the page! Nobody has altered the page while I'm editing, those alterations were done before I clicked refresh (maybe hours before!) and certainly before I started editing. The problem is that important fields are NOT REFRESHED when I click on refresh page.
Another curiosity is that I can not drag the tabs in this FF window to sort the bugzillas. I have one FF window with 30 tabs opened, each one bugzilla or related info. I try to have them sorted numerically, but FF does not allow it (today). However, if I drag some tabs to a new FF window, then I can sort them just fine in the new window, but not in the old one. This current FF window is somehow corrupted.
Changing platform in Chrome (x86-84)) to verify if FF sees it.
Just clicked refresh in FF. Platform is still "other" here, so it will cycle back when I hit submit. When you comment back, unless you tell me different, I will move this tab to a new window to see if it refreshes properly. I will not do it till you give the go ahead, in case you want some other test first.
Chrome, on refresh, sees "other". One more time, change platform again to x86-84 in chrome.
In FF, instead of the refresh button, I clicked on the bugzilla number after the edit. I see "other" in component, it has not refreshed fully.
And chrome sees all changes correctly. Thus it is demonstrated that it is a FF issue, not a bugzilla issue.
This is now reported upstream as it's unlikely openSUSE specific
Could you reproduce the problem? I see you did a test, but no comments. :-?
I could reproduce a part of the problem easily.
Thanks... I thought I had gremlins about...
This is still happening. Firefox 24.0