Perhaps I should start a new thread... What do we consider/advertise the stability/purpose of the Tumbleweed container images? It seems a bit hard to believe that this is an acceptable breakage unless they are consider toys. When I encountered this problem in a Gitlab CI runner via the message "shell not found" I assumed it was something in TW that had been updated and switch the image to Leap as I did not have the time to investigate. More recently, I dug into it (after I had to rebuild more of my TW based images) [1] and determine the following code had to evaluate to false.
elif [ -x /bin/bash ]; then
I figured there had to be something odd going as as there is no way something as fundamental as checking if a file is executable would be broken in Tumbleweed for over a month. It was not until asking in IRC someone pointed me to a bug indicating that was indeed the case. Clearly various upstream projects will make breaking changes from time to time that we have to handle as a distribution, but this seems like a very bad way to handle it. Clearly based on the bug report various openSUSE projects depend on these images working so that seems to indicate they are not considered toys. How then can we have the expectation that the entire world updates to accommodate a Tumbleweed container? I understand that ultimately that is what is needed to support the newer glibc, but is that not a reason to block the update like we do when other packages break things severely? It would be one thing if our point-release distro (Leap) was updated to handle this at the minimum and the fix had been widely available for say several years. At that point sure, point at everyone and say they need to fix it. But to roll this out knowing it will break things all over and then say things like "Not a bug." is incredible. Even running a TW container on a Leap 15.2 host this does not work. If this is acceptable for going on two months how do we expect serious usage of the Tumbleweed containers? It's great that, from what I can tell, various CI providers have patched this, but I am at a loss that the only reasonable situation for someone running an openSUSE host is to stop using the Tumbleweed container for workloads. [1] https://forums.opensuse.org/showthread.php/552485-Sometime-after-20210210-ev... -- Jimmy