SDE and Y2038

[Chris Lamb]

Hey Bernhard,

Thanks for your many MRs! I'll just close up a few things here:

> I'd prefer to have the spec limit valid values to 
> *positive* or *non-negative* integers.

Works for me. Alas, though, I think the use of SOURCE_DATE_EPOCH="0" is
basically unavoidable (despite its issues), so would support an explicit
"non-negative" (over "positive").

> I was just worried that some functions might return -1 on error that 
> could wrap around in uint64_t types (which don't make sense to use)

Nod.

> > Do you think [using SDE as a flag] warrants something in the spec or
> > just something in the docs?
>
> It caused some discussion in a few of my PRs, because I link to the 
> spec. 

Ah, very interesting. I would support the addition of such a note; we can
discuss details on the MR.

> > [..]
> perl is the easy case.

(We should make a t-shirt with this quote btw.)

ps. Funny that you mention using SDE=1 — I *used* to use that as my
dummy value, but after spending more time in -0800, I now use SDE=123123
or otherwise spam some reasonably large number in there; whatever will
prevent the local timezone rendering it as the hard-to-spot December 1969
and not the very iconic epoch. ^_^


Best wishes,

-- 
      o
    ⬋   ⬊      Chris Lamb
   o     o     reproducible-builds.org 💠
    ⬊   ⬋
      o