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
release: access fetchGit from builtins to fix parsing w/1.11 (<1.12) #1789
Conversation
Consider building the tarball, With this PR, the default If we can't make the default work for both cases, then I think this should be reverted. |
Hmm. Works for me on 1.12, don't know why it wouldn't work for you. I noticed the Nix being built in #1814 is old (cd74a55), if that's also the version you're using perhaps it's a bug that's been fixed since then? (also I don't think release.nix is meant to actually let you build "any" nix by overriding the Anyway: This PR was about 1.11 failing to even parse the release.nix, which is especially unfortunate. It will still fail to evaluate if you don't override Similarly 1.11 will fail unless you also override |
This PR simply replaces FWIW if It would be "nice" if there was a better story for use with 1.11 but since few people use |
Do you have the sandbox enabled? It fails because it cannot find git, which isn't included as a
It has not been fixed.
That is indeed unfortunate. Aside from adding |
yes I do. And I build Nix all the time on multiple machines all using sandbox ... Something strange is going on--are you saying that if you revert the commit in this PR your observed behavior re:errors in #1814 is changed? (and why is this somehow also causing you to have a missing json.h fle?? Really sounds like you're just using the wrong release.nix for your nix version?) I'd be less surprised if you found problems before/after 6d80870 , is that what you meant? Also looking at commit history around that point only reinforces the idea that your problems are mismatched release.nix for the nix source (changes to the json bits, adding git and mercurial to satisfy check reqs)-- all of which explain your problem and have nothing to do with this PR. Do you see the problem if you use the release.nix from the source you're building, or only when they're mismatched? |
I don't have access to the previous machine currently, but am now on another one, that runs Something weird is going on, and I think it is
produces
Alright. Now, let's revert this change:
And I get the same error. So, it's unrelated, right? What if we reset the change, i.e., we keep the change, but not the commit:
And it builds! Note that using master at 98f3c75 and cherry-picking my change from #1814 did not make it build on this machine. However....when I have the change in there (adding |
I will try (later) building with a newer Nix version. |
That is very strange!! I'm sorry if this was caused by this PR, but regardless this is rather unexpected. Minor observation, the failing log looks like:
Which I'll note has the git revision of ae8884b (and "revision count" is about 1000 revisions older than master revision used 98f3c75. Which leads me to believe you maybe are being bitten by the issue fixed by #1755 or a related one-- namely, if you used fetchGit on a clean git repository 'master' was used regardless of current checked out branch. I don't suppose your local 'master' reference points to ae8884b? Related: what's the version of nix being used to evaluate all of this? And is daemon the same version? Oh-- your final comment really makes this sound like the issue fixed in that PR. Fingers crossed! |
Currently nix "stable" (tested with 1.11.16) aborts when trying to process release.nix:
This PR accesses
fetchGit
throughbuiltins
so thatrelease.nix
parses,and can be evaluated with 1.11 if proper
nix
andnixpkgs
arguments are provided.