Yet more historical data for Iran
Yet another discovery, which clarifies what happened in 1977 and 1978 in Iran. This is from Kayhan newspaper, one of the major Iranian newspapers, from March 20, 1978, page 2: "Pull the clocks 60 minutes forward As we informed before, from the fourth day of the month Farvardin of the new year [=1978-03-24], clocks will be pulled forward, and people's daily work and life program will start one hour earlier than the current program. On the 1st day of the month Farvardin of this year [=1977-03-21], they had pulled the clocks forward by one hour, but in the month of Mehr [=1977-09-23], the clocks were pulled back by 30 minutes. In this way, from the 4th day of the month Farvardin, clocks will be ahead of the previous years by one hour and a half. According to the new program, during the night of 4th of Farvardin, when the midnight, meaning 24 o'clock is announced, the hands of the clock must be pulled forward by one hour and thus consider midnight 1 o'clock in the forenoon." This implies that in September 1977, when the daylight savings time was done with, Iran didn't go back to +03:30, but immediately to +04:00. I'm still trying to figure out when exactly did Iran go back from +4:00 to +03:30. I'll report back when I find it. Based on the time of true noon announced in newspapers of the time, I've triangulated it to some time between 1978-11-05 and 1979-01-15, but it looks like major Iranian newspapers were stopped from publishing during most of that time period, so I can't figure out more using this method. For now, two changes are needed in tzdata, first is the start of daylight saving in 1978, the other the switch to +4:00/+5:00, which is currently listed as Nov 1977, but should change to exactly the end of daylight saving time in 1977. Thanks again, Roozbeh
Thanks for the further data on Iran. Unfortunately the obvious patch to the "asia" file uncovered what appears to be a longstanding bug in zic.c. So I installed the attached three patches to fix this. The first patch refactors zic.c as I had trouble following its local-variable reuse, which is no longer needed with today's compilers. The second patch fixes the zic.c bug in two different ways, sort of a belt-and-suspenders thing. The third patch updates the "asia" file to reflect our latest understanding of Iran's civil time history, with a workaround to avoid the bug in zic 2022a and earlier. PS. Sorry about misspelling your name in the last commit message.
participants (2)
-
Paul Eggert -
Roozbeh Pournader