307571 - sometimes extension compatibility check does not fire on application update

Status: RESOLVED WORKSFORME

(Toolkit :: Application Update, defect)

Description

[Asa Dotzler [:asa]]

Install Firefox 1.0.6 and a few extensions. 
Install Firefox 1.5 Beta 1 candidate and launch Firefox at the end of the install. 

Results:
Some of the time the extension compatibility check doesn't come up and the app
launches. Checking the extension manager shows extensions disabled as not
compatible with 1.4.

We've seen this on all platforms and I personally see it about one in 5 attempts.

Comment 1

[Mike Schroepfer]

Rob, have you been able to reproduce this yet? 

Asa/QA can you test around this some more on the current branch builds? We fixed
several other issues related to this after beta1.

Comment 2

[Robert Strong (they/them - no direct email)]

I was able to reproduce it earlier but didn't have time to debug it... I'll take
a look a it this evening. Thanks.

Comment 3

[Robert Strong (they/them - no direct email)]

BTW: I highly suspect that the first patch in bug 307559 has fixed this issue...
QA testing would be very much appreciated.

Comment 4

[Marcia Knous [:marcia]]

tested on Mac using Mozilla/5.0 (Macintosh; U; PPC Mac OS X Mach-O; en-US;
rv:1.8b5) Gecko/20050920 Firefox/1.4 build.

1. Started with a fresh 1.0.6 profile and installed 3 extensions (2 had older
versions) - GMail Notifier, Foxytunes and Flashgot.
2. Launched the beta and got the extensions comp check, 2 of the three
extensions were shown in the checker as expected. It went out and found the
update for Foxytunes but not for Flashgot, despite the fact it was shown.
3. After relaunch, Flashgot showed as "disabled not compatible."

I can try some more scenarios - Asa said he got this 1 out of 5 times so I will
try a few more times.

Comment 5

[Moritz Strübe]

Ok here is what I did:
I installed FF 1.0.6 and started it.
After switching back to FF 1.4 the Checking-Dialog showed up. Just at that
moment my wlan connectlion disconnected. FF started without showing the Window
with the incompatibel extensions. After going back to 1.0.6 without installing
and starting FF 1.5 again it worked fine. 

Using W2k
Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.8b5) Gecko/20050920 Firefox/1.4

Comment 6

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8b5) Gecko/20050920 Firefox/1.4

I followed the same procedure as Marcia this afternoon... Reinstalled a fresh
copy of 1.0.6 ten times and got the extension compatibility *every* time I ran
this build for the first time. I made sure that I had at least three extensions
each time, but I installed  a different combination (I think) for each run.

Any more specifics on how to reproduce this one? Maybe a specific extension
could be causing this?

Comment 7

[Robert Strong (they/them - no direct email)]

I used the same extensions for testing this time as I used when I experienced
this on both Linux and Win32. I experienced it while writing a couple of patches
that have since landed and this bug may have been fixed by one of them.

With the short time frame between the version change which causes this code to
be used and the release it would be great if this additional testing continued.
This will give us a better chance of preventing a bug like bug 307235 from
making it into a beta or a release.

Comment 8

[Moritz Strübe]

Maby as additional Information, because it sound like you are creating fresh
profiles: I reproduced it with my standard profile (10 extensions or so) so it
hasn't to do with profiles which have never been to FF 1.4.

Comment 9

[Robert Strong (they/them - no direct email)]

Actually, when I was able to reproduce this I was using the same profile that I
had made a backup of. When using this profile I also experienced it
intermittently as was the case in comment #0

Comment 10

[Asa Dotzler [:asa]]

Rob, can you look into this?

Comment 11

[Robert Strong (they/them - no direct email)]

Asa, I looked into this yesterday and was no longer able to reproduce. Also, the
only reports from the callout for QA testing stated that they could not
reproduce as well - there was one comment that involved losing a net connection
causing unusual behavior but that isn't this bug. I'll keep trying to reproduce
during my other testing and if anyone can reproduce this please comment in this bug.

Comment 12

[Asa Dotzler [:asa]]

OK. If it's not reproducible, then there's no point in keeping a bug open. I was
sort of hoping someone could look at the code and try to determine why this
might have happened (or be happening) since it's not so easy to reproduce. 

Comment 13

[Robert Strong (they/them - no direct email)]

I have reviewed the code for this issue and regretfully nothing stands out.
Also, it would be great if testing of the upgrade code continued since this code
seldom gets exercised... even if this bug has been fixed inadvertently a
different bug could slip by as bug 307235 almost did.