However what I'm actually seeing is as follows. (1) Set a Linux machine to "Brazil/East". Check this using the "date" command line command.
(2) Confirm that select CURRENT_TIMESTAMP from rdb$database does report the Brazilian local time as expected. (3) Next, I've got a table with a TIMESTAMP column called TIMEDATESAVED. The BEFORE UPDATE trigger includes new.TIMEDATESAVED = CURRENT_TIMESTAMP; When I modify a record in that table I find that TIMEDATESAVED gets set to the the current UK wall clock time, *not* the Brazilian time as at (3) above. That's looking at the field both using a SELECT from isql and using Database Workbench, both giving the same result - I haven't made any attempt to see what the actual binary on the disk is. ------------------ So it seems from these experiments that CURRENT_TIMESTAMP sometimes reports the Brazilian time and sometimes the UK time. I just don't believe this. Can anyone make any other suggestion as to what is going on? On 13/01/2017 11:21, Tim Ward wrote: > On 12/01/2017 14:09, Tim Ward wrote: >> Sorry if this is a really basic question, but some time spent >> searching has failed to find for me the definitive detailed >> documentation on exactly how Firebird handles time zones (I'm >> particularly interested, to start with, in using CURRENT_TIMESTAMP to >> create a value in a TIMESTAMP field, and later reading it out and >> using it to display in a UI, but that's only to start with). >> >> Can someone point me in the right direction please? > > So to summarize the replies (thanks all): > > (1) Firebird does not attempt to do anything at all with time zones, > and has no concept of UTC. > > (2) CURRENT_TIMESTAMP and its relatives ask the operating system for > current wall clock time. > > (3) This is what gets stored and read back out of tables, with no time > zone manipulation. > > (4) So if for example CURRENT_TIMESTAMP is called when it's 0700 UTC > but 0500 local wall clock time as configured in the operating system, > then the value will be 0500 at all times and places in Firebird > (unless arithmetic is done in it by explicit application code). > > (I'm trying to understand the behaviour of an existing system here, > not design anything new.) > -- Tim Ward
