Page MenuHomeFreeBSD

zoneinfo: also install tzdata.zi and version
ClosedPublic

Authored by philip on Sun, Mar 30, 5:27 AM.
Tags
None
Referenced Files
Unknown Object (File)
Sat, Apr 5, 2:38 AM
Unknown Object (File)
Sat, Apr 5, 2:38 AM
Unknown Object (File)
Fri, Apr 4, 8:31 AM
Unknown Object (File)
Wed, Apr 2, 9:29 AM
Unknown Object (File)
Wed, Apr 2, 5:58 AM
Unknown Object (File)
Tue, Apr 1, 11:10 PM
Unknown Object (File)
Tue, Apr 1, 10:03 PM
Unknown Object (File)
Tue, Apr 1, 9:32 AM
Subscribers

Details

Summary

Build and install the shrunk version of the time zone database required
by libc++ 19 std::chrono::current_zone (etc). While here, also install
the tzdata 'version' file to align more closely with other vendors.

PR: 282377

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

philip created this revision.

This builds the "main" format of the time zone database. That includes backward compatibility links but it doesn't include historical ("backzone") data. The "redo" bit controls how leapseconds are handled. We may want this to be posix_right. I'm not sure what other vendors do here.

LGTM. We can address leap seconds etc. after commit.

This revision is now accepted and ready to land.Sun, Mar 30, 3:13 PM
dim added a subscriber: dim.

Great, thanks.

Setting it to posix_right i think is a terrible idea as the default. One that should be carefully considered in the larger backdrop. Iirc, being able to opt into posix_right can be good for some applications, but switching the default is HUGE pola and requires a systemic review for far too many use cases, many of which won't be known untils years later. Anybody using smeared leap seconds won't want it, for example. There's two, completely irreconcilable warring factions and way stronger evidence than just what other vendors are doing is going to be needed since there's about a 40s difference in what you set the system time to for starters... it shouldn't be done in phab.

There's talk of ditching leaps entirely and its unclear if we will have another one too.. .

Thanks for the feedback. That was my reasoning for doing posix_only instead of the tzdata default posix_right.

It sounds like it's currently not worth the bother of doing a build knob for changing this.

I'll put this in main as-is. I don't currently have a plan to merge this to stable. The earliest I would merge it is after a future tzdata release in the second half of this year or so. While there are no current consumers of this, this will want a long time to soak if libc++ picks it up.

This revision was automatically updated to reflect the committed changes.