Re: [musl] [bug] musl Daylight Saving Time across year bug

[Rich Felker <dalias@libc.org>]

On Tue, Dec 19, 2023 at 01:52:59AM +0000, wanghaitao (G) wrote:
> Hello everyone,
> I found a bug related to musl daylight saving time setting.
> We are set to enter Daylight Saving Time at 10pm on the last day of 2021,
> offset Daylight Saving Time by 2 hours, and end Daylight Saving Time on the first day of 2022.
> 
> However, it did not recover immediately after entering 2022, but it was delayed for more than ten hours before recovery to normal time.
> 
> We use the musl version 1.2.3
> We put the TZ variable settings in /var/time/TZ and use soft links to use it.
> 
> cat /var/time/TZ
> GMT-08:00DST-10:00:00,M12.5.5/18:45:0,M1.1.6/6:0:0
> we have a soft link from /system/etc/TZ -> /var/time/TZ, and /etc/TZ -> /var/time/TZ
> Here is our env info:
> root:/# cat /proc/version
> Linux version 5.10.97+ (z30045507@kwemhisprc00189) (aarch64-none-linux-gnu-gcc (
> Arm GNU Toolchain 12.2.Rel1 (Build arm-12.24)) 12.2.1 20221205, GNU ld (Arm GNU
> Toolchain 12.2.Rel1 (Build arm-12.24)) 2.39.0.20221210) #1 SMP PREEMPT Thu Dec 1
> 4 10:14:45 CST 2023
> 
> After looking at the src code I found in the src/time/__tz.c:406, there is a comment:
> FIXME: may be broken if DST changes right at year boundary? Also, this could be more efficient.
> 
> I would like to ask if the current bug is the FIXME written in the comments here. Is there any solution?
> 
> [cid:image003.png@01DA31CC.C3E3DA30]

I looked at this briefly when you first posted it, but then I think it
slipped through the cracks. Thanks for bringing it to our attention. I
hope to follow up on it sometime after the upcoming 1.2.5 release.

BTW, is this a real timezone you're trying to make use of, or a
constructed corner case for testing purposes?

Rich