fix: preventing args to service commands #45050
Conversation
also outputting all commands for help files closes keycloak#44975 Signed-off-by: Steve Hawkins <[email protected]>
|
@Pepo48 this reestablishes the replacement logic, but I see that more is still needed to enable the tests on windows. |
|
|
||
| import static org.keycloak.quarkus.runtime.cli.command.AbstractAutoBuildCommand.OPTIMIZED_BUILD_OPTION_LONG; | ||
|
|
||
| @WithEnvVars({"KEYCLOAK_ALL_COMMANDS", "true"}) |
There was a problem hiding this comment.
With this approach, we won't know the hiding of the OS-specific cocmmands actually works.
There was a problem hiding this comment.
Sorry, I meant to add a Linux specific test showing that these commands won't work there.
The main thought here is that we don't need helpcommanddisttest to be OS dependent.
I'd also be fine to extend this approach to a way to set the OS for picocli, rather than just relying upon the Environment method.
@shawkins correct, as I mentioned in the Windows service PR, this issue most likely has nothing to do with the changes introduced by the service feature. If the test had been enabled, it probably would have failed a long time ago. It’s a tricky one, because I couldn’t reproduce it locally. The raw GitHub logs show just a single trailing space (you can try searching the logs for "Usage: "), and I can’t directly download the received .txt files to check whether there’s anything more. I tried to implement a scrubber using various regexes, but again - what worked locally didn’t work in CI. Therefore, I suspect there must be something else at play, possibly an environmental difference or a Picocli discrepancy. As a result, simply adding the tag did not resolve the issue for me either, and it requires deeper analysis. I’m now wondering whether someone noticed this a long time ago and whether this was the actual reason the tests were never enabled on Windows runners. 😄. |
also outputting all commands for help files
closes #44975