ubuntu-phone team mailing list archive

[Tony Espy <espy@xxxxxxxxxxxxx>]

On 03/23/2015 03:37 PM, Robert Park wrote:
> On Mon, Mar 23, 2015 at 7:33 AM, Tony Espy <espy@xxxxxxxxxxxxx> wrote:
> > I noticed over the weekend that 'citrain device-upgrade' was getting the
> > silo wrong ( ie. I'd specified ubuntu silo 13, and it started to process
> > silo 11 ).
>
> I have no idea how that could even be remotely possible. Unless your
> pin is "0011" and you got your pin / silo number mixed up when
> invoking the command. Check the source for the citrain tool:
>
> http://bazaar.launchpad.net/~phablet-team/phablet-tools/trunk/view/head:/citrain#L59
>
> Unless you've discovered a bug in *printf*, there's really no
> possibility for the silonumber to not be exactly what you entered.

There error happened when the PPA was being added to the device's apt 
config, and no my PIN isn't 0011. ;)-

I'd seen this before for sure.  I'll try again later on and see if I can 
reproduce again.

> > I also noticed that the dashboard PPA link for my silo was wrong.  It's OK
> > this morning, but just wanted to give a heads-up that things are still a bit
> > unstable.
>
> This is really troubling, as the dashboard is also "too simple to
> fail". You can't just say "things are a little bit unstable", that's
> not how it works. The problems we're having are in the google
> spreadsheet, due to various issues with certain custom spreadsheet
> formulas being so inefficient that they crush google's servers.
> Neither the `citrain` shell command nor the links to the PPAs from the
> dashboard have anything to do with the spreadsheet, and as a result
> they should be totally unaffected by any spreadsheet issues, and if
> they are having issues then those are serious problems unto
> themselves.
>
> So, in the case of the PPA links in the dashboard, those are literally
> generated by a for loop. If the displayed silo name does not match the
> PPA being linked to, well, I can't even imagine how such a situation
> could happen. Your computer is behaving non-deterministically and you
> should get that checked out. Just look at the source:
>
> https://github.com/robru/citrain-web/blob/master/index.html#L61-L68
>
> There just isn't a way that the value of "siloname" can change between
> line 65 and 67.
>
>
> In short, neither of the issues you've described are possible.

Never say never...

> I recommend running memtest.

Sure, I'll run a memtest, but I doubt memory corruption would lead to 
such usable ( albeit wrong ) results.

I checked the dashboard and the PPA link was fine this morning.

Maybe sunspots.

Will let you know if I see either again.

/t