Re: Fwd: [Qt bugreports] Updates for QTBUG-75585: Massive performance regression in QML Date object
Tobias Leupold <[email protected]> Mon, 21 Sep 2020 14:40:02 +0200
| Newsgroups | gmane.comp.kde.kimdaba |
|---|---|
| Message-ID | <20798721.EfDdHjke4D@ginuog> |
Am Montag, 21. September 2020, 14:23:54 CEST schrieb Robert Krawitz: > Here's the answer I got. Well, if using only UTC dates makes it better, we could think about doing that. At the moment, we don't even include a timezone information in the XML file, e. g. we have startDate="2012-08-16T12:01:09". If this won't break downwards compatibility (I didn't look at the code yet), we could convert all timestamps to UTC, supposing they are in local timezone. Like startDate="2012-08-16T14:01:09Z" or such for the above example. This could be done via a db version bump and update procedure and would only have to be done once, processing all dates. For newly added images, we simply do like so without a noticeable performance impact, as the conversion only takes place once, when they are initially added. Actually, this is imo a shortcoming that should be fixed anyway, as the currently used format isn't unambiguous and will deliver the wrong time if a user e. g. moves to another timezone. Around midnight, it will even deliver the wrong day ... also, we have a problem with DST, as once a year, the same hour passes two times with no way to know when exactly the photo was taken without knowing the timezone!