PedroMJJun 16
daniel:// stenberg://

Today I learned that glibc has a broken %s implementation in strftime. It applies the timezone even when we want UTC...

https://github.com/curl/curl/issues/22038

%time{%s} in --write-out returns wrong epoch on non-UTC hosts · Issue #22038 · curl/curl

I did this Summary curl --write-out '%time{%s}' (and %time{%s.%f}) outputs an epoch value that is offset by the local timezone offset from UTC. On a host in Asia/Shanghai (UTC+8), the value is cons...

GitHub
Jun 16 at 7:12amWeb
Show threadBhaskar ChowdhuryJun 16
@bagder Oh, heck :(
Show threadHarry SintonenJun 16
@bagder How is this broken? It's a non-standard extension to strftime(), so it can be whatever it wants.
Show threaddaniel:// stenberg://Jun 16
@harrysintonen I would still say that applying the timezone is wrong, but sure it's just my opinion. I'm not going to insist
Show threadHarry SintonenJun 16

@bagder Man page says "The number of seconds since the Epoch, 1970-01-01 00:00:00 +0000 (UTC). (TZ) (Calculated from mktime(tm).)"

Everything there suggests timezone is applied.

Not even arguing if that makes sense, just that it is documented to do that.

Show threaddaniel:// stenberg://Jun 16
@harrysintonen I'll still claim that is wrong, but documented =)
Show threadφJun 16
@bagder @harrysintonen gnu manpages once again serving as a comprehensive bug tracker
Show threadHarry SintonenJun 16
@bagder Yeah, a lot of things are documented and still very wrong. :>
Show threadIain McLarenJun 16
@bagder ah as soon as any of my code touches UTC in any way I just know I’m about to go down a 2-3 day rabbit hole
Show threadMathias HasselmannJun 16

@imclaren Depends on programming language, and within programming language on toolkit, I'd say?

@bagder

Show threadIain McLarenJun 16
@taschenorakel @bagder respectful strong disagree. UTC is like complex concurrency. It’s a bag of hard even with beautiful programming languages.
Show threadMathias HasselmannJun 16

@imclaren UTC itself or conversion from and to local time?

While huge effort is needed to compute the official UTC timestamp, for us mere mortals it's just "Whatever number my NTP server of trust gives me", not?

Biggest issue, as with all wall times: It is not monotonic...

@bagder

Show threadIain McLarenJun 16

@taschenorakel @bagder conversions. Although I’ve never seen live code dealing with utc that doesn’t involve conversions.

Obligatory xkcd: https://xkcd.com/1883/

Supervillain Plan

xkcd
Show threadMathias HasselmannJun 16

@imclaren Yes, but this is local times, not UTC. Local times are complicated.

@bagder

Show threadIain McLarenJun 16
@taschenorakel @bagder maybe it’s just me but I have never worked on prod code that didn’t involve conversion to localtime. For me, working with utc means working with localtime
Show threadBen ZaninJun 16
@bagder oooooof, that's useful to know about. Thank you for mentioning it.
Show threadPeter KreftingJun 16

@bagder Does the refrence implementation in IANA tzdb handle it differently?

There was a discussion about it a few years ago on the tz mailing list: https://mm.icann.org/pipermail/tz/2024-January/058539.html

[tz] strftime %s