[rb-general] Q. Should a tool that implements SOURCE_DATE_EPOCH abort if it can't ensure reproducibility?

Daniel Shahaf danielsh at apache.org
Sat Jun 24 02:50:34 CEST 2017


Dan Kegel wrote on Fri, 23 Jun 2017 16:39 -0700:
> is SOURCE_DATE_EPOCH a best-effort flag, or a must-deliver kind
> of flag?
>
> Can I get a clarification on the spec from the group?

My reading:

    Upstream build processes MUST use this variable for embedded
    timestamps in place of the "current" date and time.

means:

1. If the envvar is set and the build process is unable to set embedded
   timestamps to that value, then it MUST exit with an error.

2. If the envvar is set and the build process is unable to ensure
   reproducibility for non-timestamp-related reasons, it MAY continue.
   It MAY issue a warning as well, and arguably SHOULD issue a warning
   (because reproducibility is good in general, not because S_D_E
   implies the user wants reproducibility).

None of this has any bearing on the semantics of, say, --enable-reproducible-build
flags to configure scripts; these aren't covered by S_D_E.

Cheers,

Daniel


More information about the rb-general mailing list