GNU Coreutils 9.4 Adds Experimental "--enable-systemd" Option, Faster Split

[szymon_g]

PLD/

I thought PLD was dead?

[kylew77]

Now I'm curious as to the the epoch and range of the system RTC (i.e. what keeps time when the system is turned off). Presumably those too will overflow at some point.

At least on PC I expect the protocol between the OS/BIOS/UEFI and the hardware clock to be somewhat standard. Other platforms will likely differ.

I worked for a company that used ancient Supermicro systems as part of our fleet, they had bug in either 2020 or 2021 that made it impossible to access the BIOS after the time rolled over. Took us all by surprise! These were early 64bit Xeons.

[NotMine999]

I read the GNU Release Announcement. Then I read 2 more times just to make sure I had not over looked anything.

I see the following bit:

pinky, uptime, users, and who no longer misbehave on 32-bit GNU/Linux
platforms like x86 and ARM where time_t was historically 32 bits.
Also see the new --enable-systemd option mentioned below.
[bug introduced in coreutils-9.0]​

So I went looking for that "option mentioned below." and it's buried in this statement.

On GNU/Linux platforms where utmp-format files have 32-bit timestamps,
  pinky, uptime, and who can now work for times after the year 2038,
  so long as systemd is installed, you configure with a new, experimental
  option --enable-systemd, and you use the programs without file arguments.
  (For example, with systemd 'who /var/log/wtmp' does not work because
  systemd does not support the equivalent of /var/log/wtmp.)​

Ok, so it's documented in a wordy statement. I guess it pays to read the documentation (Release Announcement in this case), but how many of us actually READ IT when we usually just SKIM IT?

[PluMGMK]

Now I'm curious as to the the epoch and range of the system RTC (i.e. what keeps time when the system is turned off). Presumably those too will overflow at some point.

At least on PC I expect the protocol between the OS/BIOS/UEFI and the hardware clock to be somewhat standard. Other platforms will likely differ.

Well, I once managed to accidentally set the system time to 2085 by calling int 1Ah without setting the registers correctly. Actually, looking at the list of functions I have no idea which one it could have been, but at least I know it's possible…

[microcode]

Do coreutils binaries have to provided in two variants, systemd and non-systemd (for systems where systemd can be used by is optional; like PLD/Linux) or did they make this in sane and compatible way, so one "binary fits all"?

Bruh if you are packaging GNU coreutils, you can compile it to be compatible with your distribution; if you aren't a distribution packager, then just build the version that will be compatible with the system you are building.

[Ironmask]

Was this ever even an issue in Windows? I thought DOS/Windows had their own threshold dates completely separate from this 2038 one…

Windows NT's time was always represented as a 64-bit integer, it's epoch is the year 1800.
Linux used a 32-bit integer well into the 2000s which is honestly psychotic.

[arekm]

I read the GNU Release Announcement. Then I read 2 more times just to make sure I had not over looked anything.
Ok, so it's documented in a wordy statement. I guess it pays to read the documentation (Release Announcement in this case), but how many of us actually READ IT when we usually just SKIM IT?

So you see that release announce doesn't explain anything.

But looking at source code and it looks that binaries build with --enable-systemd will work on both worlds (with and without systemd).

int
read_utmp (char const *file, idx_t *n_entries, STRUCT_UTMP **utmp_buf,
int options)
{
# if READUTMP_USE_SYSTEMD
if (strcmp (file, UTMP_FILE) == 0)
/* Imitate reading UTMP_FILE, using systemd and Linux APIs. */
return read_utmp_from_systemd (n_entries, utmp_buf, options);
# endif

return read_utmp_from_file (file, n_entries, utmp_buf, options);
}

[cynic]

Linux used a 32-bit integer well into the 2000s which is honestly psychotic.

we're all so good in taking decisions... 20 years later!

[Ironmask]

we're all so good in taking decisions... 20 years later!

And again, for the third time:
Windows NT had a 64-bit timestamp in 1993.
This was not revolutionary, it was actually fairly standard.
If only Linux wasn't trying to poorly mimic a hacked-together OS architecture that was designed on a PDP-11 that even it's own creator disowned.

[cynic]

If only Linux wasn't trying to poorly mimic a hacked-together OS architecture that was designed on a PDP-11 that even it's own creator disowned.

when Linus started writing Linux he had no idea where his OS would have been in 30 years.
It wasn't designed to last, and surely not to last this long.
He used a design that was working at the time and was enough for his needs.