Add compatibility with RHEL7 (and YUM v3) #45
Conversation
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
|
@janblazek you can merge this anytime. Or do you want me to do it? |
|
I think this may have broken all installability tests for Fedora PRs, e.g. http://artifacts.dev.testing-farm.io/ef9fb00a-212e-4bc9-a51b-5fc1c9b945a6 from https://src.fedoraproject.org/rpms/anaconda/pull-request/151# . It changes the command used to find packages from This doesn't seem to be affecting the installability test run on Koji package builds, at least I see many passed results for that version of the test. I guess dnf-utils happens to be present in that environment, or something. |
fedora-ci#45 changed two out of three places we previously called `"$YUMDNFCMD" repoquery` to instead just call `repoquery` (it missed one). This is because `yum repoquery` doesn't exist on EL 7. However, it's also not really safe to assume the standalone `repoquery` command will always be available either; it's part of yum-utils or dnf-utils and there's no reason to assume that will always be present. This adds a fallback, and consistently uses the command we determine across all three locations. Signed-off-by: Adam Williamson <[email protected]>
|
Additionally, it seems to have missed changing one I checked and confirmed that Weirdly, a run on a similar PR three hours later worked: https://artifacts.dev.testing-farm.io/a221e399-1a3a-419f-bacc-39e6a53988e7/ from https://src.fedoraproject.org/rpms/anaconda/pull-request/152 . I can't see what the difference between the two runs was. Did anyone do anything intentional to fix this? You can tell the From the passed test: that box is populated with the content of I came up with two alternative options to fix this, but seems like it might be worth figuring out the mystery first... edit: oh, I see, the failed test got mini-tps-0.1-171.fc41.noarch.rpm but the passed test got mini-tps-0.1-170.fc41.noarch.rpm (despite running later). Maybe someone saw this problem and rewound the COPR, or something? |
fedora-ci#45 changed two of the three places we try and do a repoquery operation from calling `"$YUMDNFCMD" repoquery` to just calling `repoquery`, but it missed one. This fixes the remaining one. Also, this attempts to ensure the plain `repoquery` command will always be available. It's not safe to assume this, because it's part of dnf-utils (on Fedora) and yum-utils (on EL), it's not part of core dnf/yum, and nothing much else requires it. It does seem to be missing in the environment in which tests on Fedora package pull requests run, for instance, which I think causes the test to fail on all pull requests at present. Signed-off-by: Adam Williamson <[email protected]>
…i#47) fedora-ci#45 changed two of the three places we try and do a repoquery operation from calling `"$YUMDNFCMD" repoquery` to just calling `repoquery`, but it missed one. This fixes the remaining one. Also, this attempts to ensure the plain `repoquery` command will always be available. It's not safe to assume this, because it's part of dnf-utils (on Fedora) and yum-utils (on EL), it's not part of core dnf/yum, and nothing much else requires it. It does seem to be missing in the environment in which tests on Fedora package pull requests run, for instance, which I think causes the test to fail on all pull requests at present. Signed-off-by: Adam Williamson <[email protected]>
…ora-ci#47) fedora-ci#45 changed two out of three places we previously called `"$YUMDNFCMD" repoquery` to instead just call `repoquery` (it missed one). This is because `yum repoquery` doesn't exist on EL 7. However, it's also not really safe to assume the standalone `repoquery` command will always be available either; it's part of yum-utils or dnf-utils and there's no reason to assume that will always be present. This adds a fallback, and consistently uses the command we determine across all three locations. Signed-off-by: Adam Williamson <[email protected]>
|
well, this time @jpopelka just deleted the broken build from COPR, which caused subsequent tests to use an older build. |
Related: https://issues.redhat.com/browse/OSCI-5194