Update not working properly (by design)

Mark Banner mbanner at mozillamessaging.com
Thu Mar 10 15:35:32 UTC 2011

On 09/03/2011 12:59, Ben Bucksch wrote:
> I understand the test matrix concern. Why can't the updater, when it 
> sees that the latest version is another major branch, just fetch a 
> full package (no binary diff) and install that? I thought it already 
> does that, and if so, I don't understand the problem. Testing-wise, 
> it's equal to a new download.
It isn't as equal as you imply:

We can post a new download on the website at any time, we know the 
download works because we can click the link and verify a file is 
offered and saved. We also know the installer works because we've 
already tested it before we shipped it.

When we post a new update, there are various things that could go wrong: 
pointing to incorrect locations, checksums incorrect, incorrect pushing 
to an update channel - at least two of those have happened at times in 
the last year. So currently we have to verify the update actually works 
- from beginning version to end version.

Whilst we have some automation, generating and testing updates for all 
possible past releases could be quite expensive (in the range of up to 
an extra day per release).

I think there may be some less expensive things we can try as well - 
e.g. putting a big warning on the what's new page for older releases 
that these are now out of date and the user should update. Whilst that's 
not necessarily quite the same, it still gives the user a fair warning.

So whilst I can see it might be better for the small amount of users 
that don't update regularly, I think that it would be a significant 
amount of work for where we currently are. We may be able to do this 
easier in future with our automation progressing forwards; and I'll be 
discussing where things our with the Firefox folks next time I see them. 
I'm just not convinced that now is the right time for us to be spending 
the necessary effort to do this - in both updating our systems to cope, 
and spending the extra time per release.


More information about the tb-planning mailing list