Error from solver: EPIPE "write"

[patricoferris]

Seems I can't get our CI unstuck on ocaml-rtree as the solver keeps failing: https://ocaml.ci.dev/github/geocaml/ocaml-rtree/commit/611d25bb6a303931c245ec69a463a871c118352b/variant/%28analysis%29

	2023-10-05 20:51.11: New job: Analyse
2023-10-05 20:51.11: Checking out commit 611d25bb. To reproduce:
                       git clone --recursive "https://github.com/geocaml/ocaml-rtree.git" && cd "ocaml-rtree" && git fetch origin "refs/pull/22/head" && git reset --hard 611d25bb
2023-10-05 20:51.11: Exec: "cp" "-a" "--" "/var/lib/ocurrent/var/git/ocaml-rtree.git-f1e70cb14ba1d0820d4be9356e6f434c9aa4d49c1205a25791f6ae6257af1bbf/.git" 
                           "/tmp/git-checkout3fdf4bc"
2023-10-05 20:51.11: Exec: "git" "-C" "/tmp/git-checkout3fdf4bc" "submodule" 
                           "deinit" "--force" "--all"
2023-10-05 20:51.12: Exec: "git" "-C" "/tmp/git-checkout3fdf4bc" "reset" 
                           "--hard" "-q" "611d25bb6a303931c245ec69a463a871c118352b"
2023-10-05 20:51.12: Exec: "git" "-c" "protocol.file.allow=always" "-C" 
                           "/tmp/git-checkout3fdf4bc" "submodule" "update" 
                           "--recursive" "--init" "--no-fetch"
2023-10-05 20:51.12: Found OCamlFormat version '0.26.0' in dotfile
2023-10-05 20:51.12: Solving with opam-repository commit: https://github.com/ocaml/opam-repository.git#refs/heads/master (4fec1114629a054deee2e73a113727b0ef5cbf9f)
2023-10-05 20:51.12: Waiting for worker…
Building on arm64-jade-1.equinix.ci.dev-10
2023-10-05 20:51.12 [INFO] Solving for rtree.dev
2023-10-05 20:51.38: Job failed: Error from solver: Failed: Unix.Unix_error(Unix.EPIPE, "write", "")

[talex5]

@mtelvers it looks like the old solver is running again on jade-1. It's logging a message that doesn't exist in the new version, and its size matches the old service binary:

$ strings /usr/local/bin/solver-service | grep "Killed by signal"
Worker %d finished. Killed by signal %a@.
 finished. Killed by signal 

It looks like the "live" branch didn't get updated after the merge (note: the test deployment was done from the "staging" branch).

[mtelvers]

@talex5 I have pushed to live and updated the workers.

[talex5]

Thanks - @patricoferris does that help?

[patricoferris]

It seems this might have just stopped OCaml-CI from working entirely? I'm just getting 500 errors ?

[talex5]

I think that's unrelated to this issue. Looking at the ocaml-ci log, I see things like:

Failed to get head of ocaml/opam-repository:refs/heads/master
[...]
GitHub API error: 403 Forbidden -- Resource not accessible by integration

This means that the analysis job can't run. Separately, the web UI seems to handle that by returning a 500 internal error:

07.10.23 11:14:36.284       dream.log  INFO REQ 68411 GET /github/ocaml-multicore/eio/commit/69c622fa0184606164bfbb6a257768c87d2a9d29/variant/%28analysis%29 10.0.1.4:55160 Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Firefox/102.0
07.10.23 11:14:36.330                       REQ 68411 Internal server error: Failed: No job for variant "(analysis)" yet