Possible inconsistency with use of TZ to change zone output '%Z"

Possible inconsistency with use of TZ to change zone output '%Z"

[Jim]

[-- Attachment #1: Type: text/plain, Size: 1321 bytes --]

1) using parameter expansion flag '%'
2) zsh/stat module
3) zsh/datetime module - strftime

TZ=UTC in all cases

All methods in outputting a time string appear to use strftime. zsh/stat
doesn't say so in man page, but uses the same formatting.

TZ is not treated the same way in all three uses. Parameter expansion
and zstat need TZ to be exported to give the correct result. Whereas
strftime does not.

1) Using parameter expansion flag '%' to get current date/time
module zsh/datetime  loaded Yes  DT=${(%):-%D{%Y-%m-%d_%H:%M:%S.%N_%Z}}
TZ exported NOT  2022-01-06_17:08:34.957909030_CST
TZ exported Yes  2022-01-06_23:08:34.958007858_UTC

module zsh/datetime  loaded NOT  DT=${(%):-%D{%Y-%m-%d_%H:%M:%S.%N_%Z}
TZ exported NOT  2022-01-06_17:08:34.958192556_CST
TZ exported Yes  2022-01-06_23:08:34.958318194_UTC

2) Output from zstat:
TZ exported NOT  2021-07-25_13:53:11.000000000_CDT
TZ exported Yes  2021-07-25_18:53:11.000000000_UTC

3) Output from strftime:
TZ exported NOT  2021-07-25_18:53:11.000000000_UTC
TZ exported Yes  2021-07-25_18:53:11.000000000_UTC

Can anyone tell me why TZ needs to be exported for 1 and 2, but not 3?
Personally I prefer not to have to export TZ to get the desired result,
but that's just me. I don't know the reasoning with each.

Thank in advance for any info you can offer,

Jim

[-- Attachment #2: Type: text/html, Size: 2509 bytes --]

Re: Possible inconsistency with use of TZ to change zone output '%Z"

[Bart Schaefer]

There's a lengthy discussion of this in the zsh-users thread with
subject "print builtin preceded by parameter assignment" from April
2019.

The upshot is that some system libraries only use the environment
strings when interpreting the time zone.

Re: Possible inconsistency with use of TZ to change zone output '%Z"

[Jim]

[-- Attachment #1: Type: text/plain, Size: 1438 bytes --]

Unfortunately my searches never found that mail thread. Sorry about that.
I'm sure the print builtin has nothing to do with the results I am getting.

1) DT=${(%):-%D{$Fmt}}
2) zstat -F $Fmt -A A +mtime $F
3) strftime -s DT $Fmt $ET

In each case the results were placed in a parameter then printed if and as
needed.

The thread left me somewhat confused, quite a bit to digest. Probably
because I'm
not a "developer" and know enough C to be a danger to myself and those
around
me. I understand that for a particular system the libraries may differ from
other systems, and once compiled, will "fix" the way it acts(environment
strings
or not). But if parameter expansion is using zsh's strftime shouldn't it
act the
same way as calling strftime directly. My results indicate that they don't.
It seems logical to me that they would. Admittedly I don't know the
internals, so
I'll accept the results and export TZ as needed.

Z-shell is still my all time favorite shell. Never cared for the de facto
linux default.
Sorry for the personal opinion.

Appreciated you taking the time to respond,

Jim

On Thu, Jan 6, 2022 at 6:58 PM Bart Schaefer <schaefer@brasslantern.com>
wrote:

> There's a lengthy discussion of this in the zsh-users thread with
> subject "print builtin preceded by parameter assignment" from April
> 2019.
>
> The upshot is that some system libraries only use the environment
> strings when interpreting the time zone.
>

[-- Attachment #2: Type: text/html, Size: 2088 bytes --]

Re: Possible inconsistency with use of TZ to change zone output '%Z"

[Bart Schaefer]

On Sat, Jan 8, 2022 at 7:33 AM Jim <linux.tech.guy@gmail.com> wrote:
>
> Unfortunately my searches never found that mail thread. Sorry about that.
> I'm sure the print builtin has nothing to do with the results I am getting.

I didn't mean to imply that you should have found that thread on your
own, sorry.  The significant bit is that there is some discussion of
possible patches in that thread.

> But if parameter expansion is using zsh's strftime shouldn't it act the
> same way as calling strftime directly.

The %-replacements for zsh's strftime do not map 1-to-1 onto every
underlying strftime implementation, so it's not a direct pass-through.
However --

> 1) DT=${(%):-%D{$Fmt}}
> 2) zstat -F $Fmt -A A +mtime $F
> 3) strftime -s DT $Fmt $ET

-- the patch discussion from the other thread comes down to:  In the
third case, the strftime builtin has an execution context in which it
is able to implicitly create a temporary export of TZ and then delete
it again before returning, so it does that (this was a change
implemented some time after the datetime module was first created).
In the first case there is no such context so a solution is messy, and
in the second (zstat) case the equivalent export has simply never been
implemented even though it would be possible to do so.