Janneke Nieuwenhuizen <[email protected]> writes: > So my question is: Do we indeed advise not to use --mtime=0 for source > tarballs? And if so, why is that?
HP-UX 'make' treats mtime 0 as a missing file, and POSIX encourage it: https://lists.gnu.org/archive/html/bug-gnulib/2024-12/msg00209.html https://lists.gnu.org/archive/html/bug-gnulib/2024-12/msg00227.html If you want to depend on GNU make behaviour, all is good. I think the bigger concern is make-based dependency build tracking of generated files inside the tarball, for 'make dist' tarballs: https://lists.gnu.org/archive/html/bug-gnulib/2024-12/msg00199.html https://mail.gnu.org/archive/html/bug-gnulib/2025-02/msg00166.html https://lists.gnu.org/archive/html/bug-gnulib/2024-12/msg00182.html I think these are weak arguments, and I suggest we experiment with --mtime=0 to see if anything breaks that cannot be fixed. Including generated files inside a tarball is a bad idea. Use git-archive tarballs, and only put source code files in git. However, we will have to live with 'make dist' tarballs for a long time, and sometimes vendoring some external files is reasonable to avoid really complicated build dependencies. So maybe Bruno's vc-mtime can be a solution here, but it is quite complex. /Simon
signature.asc
Description: PGP signature
