Publish circle-interpreter compiler debian package

[seanshpark]

circle-mlir uses circle-interpreter tool for value verification of conversion results.
for some Ops, circle-interpreter also needs update, while working with that Op in circle-mlir.

as of current workflow, these two projects have independent build/test flows.
if we have some nightly publish of circle-interpreter or one-compiler,
circle-mlir can use it for verification step.

so nightly publish of Debian package of circle-interpreter or one-compiler is needed.

---

plans

1/ enable end-to-end validation with circle-interpreter from Workflow Artifact
2/ produce circle-interprete Debian package to launchpad.net
3/ revise validation to use Debian package

[seanshpark]

bash script that I used to make a zip file

#!/bin/bash

mkdir -p bin
cp -v ../install/bin/circle-interpreter ./bin/.

cp -v ../install/lib/libloco.so ./bin/.
cp -v ../install/lib/libluci_env.so ./bin/.
cp -v ../install/lib/libluci_import.so ./bin/.
cp -v ../install/lib/libluci_interpreter.so ./bin/.
cp -v ../install/lib/libluci_lang.so ./bin/.
cp -v ../install/lib/libluci_logex.so ./bin/.
cp -v ../install/lib/libluci_log.so ./bin/.
cp -v ../install/lib/libluci_plan.so ./bin/.
cp -v ../install/lib/libluci_profile.so ./bin/.

cd bin
rm -f circle_intp_2204.zip
zip circle_intp_2204.zip circle-interpreter *.so

zip size is about 1.6M

[seanshpark]

Alternate simple solution that I can think of to solve this;

• prepare a Workflow to build circle-interpreter zip file and upload as Artifact
• manual trigger the Workflow
• get Artifact download URL
• update persistant place with the URL
  • like this comment
• in run-circle-mlir-build
  • get download URL from persistant place
  • download circle-interpreter zip

How to migrate to better place

• prepare that better place for circle-interpreter
• add new job to publish to new place
• manual check with new publish and download
• in run-circle-mlir-build
  • use new download interface to download circle-interpreter
• deprecate old job

[seanshpark]

https://github.com/Samsung/ONE/actions/runs/13363096215/artifacts/2600977892

[seanshpark]

https://github.com/Samsung/ONE/actions/runs/13363096215/artifacts/2600975402

[seanshpark]

Experiment:

• use actions/download-artifact@v4 to download Artifact
• looks like giving run-id with previous upload-artifact didn't work
  • Workflow where to download as of now: https://github.com/Samsung/ONE/actions/runs/13363096215
  • with or without github-token: with my personal token

log

Run actions/download-artifact@v4
  with:
    run-id: 13363096215
    merge-multiple: false
    repository: Samsung/ONE
/usr/bin/docker exec  2ce6cc196cc8a31db759a659341fc5603bdab18939435ff1f6cbccc4791bb12c sh -c "cat /etc/*release | grep ^ID"
Found 0 artifact(s)
No input name or pattern filtered specified, downloading all artifacts
An extra directory with the artifact name will be created for each download
Total of 0 artifact(s) downloaded
Download artifact has finished successfully

[seanshpark]

Next experiment

• download with curl / wget
• URL: `https://api.github.com/repos/Samsung/ONE/actions/artifacts/artifact_id/zip
• Workflow cannot read given secrets.BLA and drops 401 error with Bad credentials
• using curl in my local bash console, download is OK

maybe, cause this is DRAFT ?
or maybe I don't have authority, which I don't know, with Samsung/ONE secrets ?