* Jochen Sprickerhof <jspricke@debian.org> [2025-08-29 12:16]:
* Cyril Brulebois <kibi@debian.org> [2025-08-29 11:31]:Roland Clobus <rclobus@rclobus.nl> (2025-08-29):I'm wondering whether the fakeroot package is actually required. As I wrote in the initial text, I did not need to have fakeroot available, and could prevent calls to fakeroot by setting the environment variable ROOTCMD to a space (as an empty value would still set it to 'fakeroot'). As I understand it, invoking 'chroot' requires being root (it is in /usr/sbin), so within the chroot-environment one is already root and therefore does not need fakeroot on top. But there might be some scenarios, that I'm unaware of, that invoke the scripts of the installer differently and therefore the package fakeroot is indeed a requirement.I'll let Jochen comment on that, who's apparently more familiar with those things than I am. Speaking of which, the initial R³ addition (which is modified by !68) happened by applying a suggestion from Niels.build/Makefile used mknod which needs (real) root or fakeroot (that tracks the faked device files). So when Niels tested setting R³ src:debian-installer did not depend on fakeroot and the package FTBFSed. The simple solution was to set R³ to binary-targets so dpkg-buildpackage runs with fakeroot. But actually only build/Makefile needs it so we can drop the binary-targets again.
I read #1091668 now and added --owner root:0 --group root:0 to tar, as Niels suggested. I have also adopted a patch from !38 make the mtimes in the mini.iso reproducible. The only difference are now in the mini.iso files, though diffoscope says the file content is the same. Happy to get any pointers there but would be great to get !68 merged already.
Cheers Jochen
Attachment:
signature.asc
Description: PGP signature