Re: reproducible built files

[Simon Josefsson]

Paul Eggert <eggert@cs.ucla.edu> writes:

> On 12/28/24 04:38, Simon Josefsson via Gnulib discussion list wrote:
>> The tarball use the same
>> mtime for all files.
>> So I don't understand this concern.  What problem do you see if you
>> don't perform those steps?
>
> I must be missing something because for many tarballs it's obvious
> that one shouldn't use the same mtime for all files.
>
> The main reason I don't set all timestamps to 1 (0 doesn't work
> because some 'make' implementations incorrectly treat 0 as meaning the
> file is missing!) in distribution tarballs is that some 'make's treat
> equal timestamps as meaning the destination is out of date and needs
> to be rebuilt. (Although POSIX does not require this 'make' behavior,
> it encourages it.)

Which 'make' behave like that?  GNU make doesn't rebuild a dependency if
the mtime is the same as the source.  I think I've made some release
tarballs using --mtime and nobody complained, but maybe everyone tested
with GNU make (or its behaviour is now common).  I'm about to push out a
new release of 'libidn' and I'm inclined to try --mtime (as well as
--format=posix) to see if anyone notices.

/Simon

signature.asc
Description: PGP signature