Notebook checkpoint time is displayed an hour out

[takluyver]

e.g. I see "Last Checkpoint: Jun 03 09:49 (autosaved: Jun 03 10:49)", having just saved it at 10:49 local time. The notification bubble when I hit save also showed 9:49 instead of 10:49.

I don't know whether it's forcing the checkpoint time to UTC, or just somehow missing daylight saving time.

[Carreau]

Weird, they are all consistent here.

[takluyver]

Just checked, and I see the same running directly from master, so it's not
something introduced by the packaging in Julian's PPA.

On 3 June 2013 13:38, Matthias Bussonnier notifications@github.com wrote:

Weird, they are all consistent here.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/3396#issuecomment-18838118
.

[Carreau]

Running on localhost ?

[takluyver]

Yep. And Ubuntu 13.04, in case that's relevant.

On 3 June 2013 14:09, Matthias Bussonnier notifications@github.com wrote:

Running on localhost ?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/3396#issuecomment-18839725
.

[minrk]

Interesting. Maybe the time conversion is missing daylight savings or something?

[takluyver]

Ah, I think I've worked this out. You're both using Chrome, right?

We pass an ambiguous string to the date constructor. Trying new Date("2013-07-03T14:38:13.096503"); in each browser, Chrome appears to interpret it as UTC, and Firefox takes it as a local time. If I append 'Z' or '+0000' to the string to specify the time zone, they both get it right.

Dealing with that properly on the Python side looks awkward - even though we use utcfromtimestamp, it still creates a 'naive' object without timezone info. I'd be inclined to just hard code the Z into the format string in jsonutil, and specify that all dates serialised from the Python API are assumed to be UTC.

[minrk]

On the Python side, we are aiming to produce UTC every time, so that makes sense.