Closed Kenneth-Sills closed 4 months ago
Apologies, my original PR contained unsigned commits. I've rebased the branch with signed commits, so should be good now.
Bump, so this doesn't get lost.
@dmurphy18 This was not fixed by #1987. The offending line is still present. I tested and confirmed it still generates the same URL and, subsequently, the same error:
Hm... Scratch that, that URL didn't look right (missing the /salt/
your changes added) so I threw together a brand new, from-scratch environment and pointed it to develop
. That seems to work! 🎉
Closing this superseded PR.
This reverts commit e05d70558f7aa1bc4b3ecce530e1ad60f4bab7ac, reversing changes made to 59af6ad185ec7334b4eedb4fa6fa287e06956427.
What does this PR do?
I spoke a bit about what I think happened here, but in short:
The two lines of link changes in #1983 caused a regression by updating the legacy repository installation codepath with links to the new, incompatible onedir file hierarchy.
http://repo.saltproject.io/py3/debian/11/amd64/latest/salt-archive-keyring.gpg
http://repo.saltproject.io/salt/py3/debian/11/amd64/SALT-PROJECT-GPG-PUBKEY-2023.gpg
http://repo.saltproject.io/py3/debian11/amd64/latest/SALT-PROJECT-GPG-PUBKEY-2023.gpg
Subsequently, the release of v2024.04.03 has broken all the Debian 10/11 VM provisioning at my workplace (currently pinned to salt version 3004.2).
I've chosen to do a simple revert since the original PR changed code that was unrelated to the quoted issue.
__install_saltstack_debian_repository
was updated instead of__install_saltstack_debian_onedir_repository
, but the former has no ability to generate the problematichttps://repo.saltproject.io/salt/py3/debian/12/amd64/latest/salt-archive-keyring.gpg
link they were referencing without a custom repo URL being passed in with-R
.What was more likely is that the latter function attempts to download from legacy before trying the onedir repository before continuing to the
SALT-PROJECT-GPG-PUBKEY-2023.gpg
case, and the author was just seeing that uncaptured output go to STDERR and assumed it was a failure.This is supported since we never actually replicated their reported issue and they never provided evidence on the bootstrap process failing. In fact, right after the error line in their terminal we see "Hit: ...", indicating that execution continued. And looking at #1988 we can see exactly that, resulting in a successful bootstrapping. The issue there really being that the script should more clearly explain that it's a soft failure and that it's moving on to a different method OR send that initial command's STDERR to the void.
What issues does this PR fix or reference?
Resolves #1986 (closed, but not yet addressed).
Previous Behavior
All legacy repository (non-onedir) repository installations on Debian fail due to an invalid URL used to download the GPG key.
New Behavior
It works.
This does NOT address the issue described by the original PR or #1988. The STDERR line is still generated.