FormatBlock.Date.ISO8601 does not take offset into account

[megri]

I just noticed for the first time that my friend outputs the incorrect timestamp.

I'm in UTC+02:00 so at 12:00 UTC my local time is 14:00, and yet this formatter outputs "14:00Z". Basically the offset is not deducted from the timestamp before printing.

[megri]

Noticed two things:

• This is a transitive issue from perfolation
• Setting java.util.TimeZone.setDefault(java.util.TimeZone.getTimeZone("UTC")) gives me the desired behaviour, but as a JVM-wide global it's quite scary to set.. 😬

[megri]

A better solution perhaps would be to add the zone offset at the end of the string according to ISO8601-standards 🤔

[darkfrog26]

Can you clarify which log formatting you are using?

[darkfrog26]

Also, are you logging on JVM, JS, or Native?

[megri]

Can you clarify which log formatting you are using?
Also, are you logging on JVM, JS, or Native?

Using FormatBlock.Date.ISO8601on JVM, although I'm assuming the outcome should be identical across all platforms.

[darkfrog26]

It should be, but the implementations are different.

[darkfrog26]

Correct me if I'm wrong here, but the problem isn't Perfolation; it's that I'm hard-coding Z on the end of the formatted stamp when it's local time. Would it be acceptable to just fix the suffix to output the proper timezone? Then, if you want to use UTC you can set the default timezone, or just leave the default and it will be the machine's local zone.

[darkfrog26]

my proposed solution is just to replace "Z" in FormatBlock:74 to be ${l.t.Z}

[darkfrog26]

@megri please let me know if this solution is acceptable to you.

[megri]

@darkfrog26 I wasn't paying attention to this. I'm sorry. Fixing the output to use the proper timezone would definitely be preferable.

my proposed solution is just to replace "Z" in FormatBlock:74 to be ${l.t.Z}

That would be great. Note I tried reading up on how to properly format this but my reference actually doesn't seem to acknowledge this format. It's definitely better than hard-coding a Z though.

As an aside I tried using ${l.t.z} to get a zone offset — which IS mentioned in the reference — but it seems the implementation is broken as I got output looking like 2024-05-02T14:09:43 +-2-120 [INFO] <empty>.JGit.JGit:37 - this is a test!

[megri]

This works though:

 val z =
    val ms = l.t.timeZoneOffsetMillis
    val sign = if ms < 0 then "-" else "+"
    val HH = Math.abs(ms) / 3600000
    val MM = Math.abs(ms) / 60000 % 60
    f"$sign$HH%02d$MM%02d"

[darkfrog26]

Hmm, I'll have to take a look and do some testing. Feel free to create a PR here: https://github.com/outr/perfolation if you're interested in helping fix it.