I understand what you are saying and I think you are right. Thanks Ulli.
If it doesn't bother you and Eric, can you confirm that it returns the instruction from my initial query?
In theory, this should be a correct solution for my purpose of subtracting a day from a date.
But my problem is that in practice (and I have tested it on ESP8266 and ESP32, with Annex 1.43.3), it returns the date minus two days, not just one.
However, it works for Cicciocb.
That's why I asked more people, from different countries, to run the code and tell me the result.
I try to find out what my mistake is
there's probably something wrong with 'dateunix' in connection with a timezone setting.
Maybe the intermediate variable precision is insufficient, cutting away a few hours and sliding into the previous day. Hopefully ciccio can say more to this.
I checked 'unixdate$', it works correctly.
With no timezone defined it seems to work correctly, strange.
Ulli, you got it right.
If I leave the "Time Zone" box blank and reset the system, unixDate$((dateUnix(date$) - 86400)) correctly calculates the previous day.
However, as is logical, now the instruction time$ returns a value 2 hours less than that of my time zone.
snake eating its own!
But I appreciate it, because we have already found out the cause. It was driving me crazy.