-
Notifications
You must be signed in to change notification settings - Fork 1.3k
fix nightly release; more flexible reusable build #1787
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
Signed-off-by: Avi Deitcher <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
|
Thanks @cmuellner ; I am sure CI will take up to an hour. I will post here when it is done. |
|
Hey @cmuellner since this is running already, any objection to adding some more parameters to the reusable build that could get passed to |
I'm fine with that. |
Signed-off-by: Avi Deitcher <[email protected]>
|
I am like that as well. As long as the default remains unchanged ("don't break assumptions without a really good reason") and it doesn't add unusual complexity, sure. Coming momentarily. |
|
Done. Let CI run its way. |
|
Runners seem to be backed up. |
|
Some passed well, others in progress, others queuing. I am watching. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
|
Nightly build works again: https://github.com/riscv-collab/riscv-gnu-toolchain/releases/tag/2025.10.18. Thanks, @deitch, for reworking the workflows and the follow-up! |
|
Excellent! I am glad it worked out. And only one PR and one correction, when dealing with Actions, actually is pretty good. I think we did well here. |
Not an intention, as much as a hope. I am not sure why the automated emails do this. |
|
It appears (I would like confirmation) to be part of how GitHub email works. As soon as a release is created, the email goes out. The fact that you upload assets right afterwards - which you have to do, since you cannot upload assets to a non-existent release - is immaterial to it. I do wonder if creating the release as a draft, then uploading assets, then actually releasing it, would change the behaviour. Bit of a pain though. The first step would be what we do now, just changing It is worth an experiment; see if it works. |

I did three things in this PR.
artifact-nameparameter to the reusable build workflow as optional. Artifact names need to be unique within a workflow, not across distinct workflows. We only needed it because the build workflow has the main build, and sim, and multilib, etc. But for nightly, where there is only one build (matrixed across variant os, lib, arch, etc.), we don't need it. So same functionality if provided, but perfectly fine if not.riscv.tar.xz, and then we saved it in an artifact namedriscv32-newlib-gcc(etc. etc), making it unique. However, if you pulled it down (elsewhere in the workflow or to your desktop or server), it would extract all of them toriscv.tar.xz, clobbering each other. Instead, right before uploading the artifact, renamesriscv.tar.xztoriscv32-newlib-gcc-whatever.tar.xz, so that the artifact name and content are unique, and ensures no clobbering.riscv32-newlib-gcc-whatever, not-buildor-nightlyat the end, which lets us download them all to a single directory, which lets us upload them easily as part of the release creation. Also did it in an accessible directory, rather than/mnt/artifacts.