python datetime .. grrr

Larry Doolittle larry at
Fri Feb 17 05:01:48 UTC 2023

Chris -

On Mon, Feb 13, 2023 at 11:20:06AM -0800, Chris Lamb wrote:
> Larry Doolittle wrote:
> > It's timezone handling in python3 datetime.
> >     doc_now = datetime.fromtimestamp(int(os.environ["SOURCE_DATE_EPOCH"]))
>                 ^^^^^^^^^^^^^^^^^^^^^^
> Simply replacing this with "datetime.utcfromtimestamp" will make
> this reproducible.

Nice.  Thanks for pointing this out!  That explains why I couldn't fix
the problem by fussing with the following .strftime() call.

Verilator upstream source has moved on, with help from Vagrant and me,
to (mostly) ignore the time-of-build, using instead the in-built
release date or a timestamp from git.  See commit bc6a7787.

Since I didn't get reproducibility right on my first attempt, could
someone with access to r-b infrastructure check if the aforementioned
patch _actually_ results in reproducible verilator Debian packages?
If so, I can turn in a proper Debian bug report.

  - Larry

More information about the rb-general mailing list