On 21/03/2020 18:53, Eduard Bloch wrote: >> In reply to the patch at https://bugs.debian.org/954437 > I cannot accept your patch as-is. Because it's brute-force, a directory > is not neccessarily a non-volatile file. Can you give me any hints how > to distinguish between volatile and non-volatile directories? Maybe the > same as it's done for for other d-i files, by regex, so > > "|/dists/.*/installer-[^/]+/[0-9][^/]+/images/.*" // d-i stuff with revision > > means frozen contents and > > "|/dists/.*/installer-[^/]+/[^0-9][^/]+/images/.*" // d-i stuff but not containing a date (year number) in the revision directory (like "current", "beta", ...) > > is volatile? (those regexps are already defined, I just need to change > your patch to make them applicable to directories) Agreed. The patch is a brute-force hack, but at least it allowed me to build my first CD unofficial images. After I've written my summary to the list, I dug deeper into live-wrapper. It uses either the released installer (using the apt proxy), which doesn't contain versioning information in the path, but will most probably be rather non-volatile, from https://deb.debian.org/debian/dists/buster/main/installer-amd64/current/images/cdrom/ or the daily build directly (without apt proxy) from https://d-i.debian.org/daily-images/%s/daily/cdrom/ However, I currently think that my issue is located within the live-wrapper package. lwr performs a download which isn't used. In lwr/utils.py the line check_url(base_url) should be removed. Then I wouldn't need a patched version of apt-cacher-ng at all... > Also, what's your deadline? I was planing to release a new version of > ACNG in a couple of days anyway. Do you also need it in backports? If you still would like to include the patch, I think that the content listing of a directory should always be considered volatile. But, as said, I think that the bug should be reassigned to the live-wrapper package. With kind regards, Roland Clobus
Attachment:
signature.asc
Description: OpenPGP digital signature