Fix ci #341

fghoussen · 2022-02-06T11:05:20Z

Pull request purpose

fixing CI that seems broken again... help would be appreciated here...
Top priority PR: no CI (arms), no merge (chocolate)...

fghoussen · 2022-02-07T12:42:07Z

CentOS seems definitely dead! @sylvestre : kill? Or try to replace it with RedHat https://faun.pub/how-to-build-rhel-based-docker-images-on-redhat-enterprise-linux-def012f84890?

sylvestre · 2022-02-07T12:50:10Z

yeah, red hat sounds great

fghoussen · 2022-02-09T20:41:07Z

basic redhat podman test: https://access.redhat.com/solutions/3696691

fghoussen · 2022-02-10T08:42:11Z

Are Github actions broken? Commits are not seen on opencollab?!...
https://github.com/opencollab/arpack-ng/actions/runs/1820338627 => opencollab side: centos should be deleted ?!
https://github.com/fghoussen/arpack-ng/actions/runs/1820338492 => my side: redhat test OK

What's going on? Any help/clue is appreciated here

fghoussen · 2022-02-10T12:58:40Z

Brilliant?!... pull_request does not see secrets anymore, so docker jobs are down...

fghoussen · 2022-02-10T17:15:42Z

On PR, commit is stuck at master: PR'ed code is not tested...

fghoussen · 2022-02-12T11:07:38Z

OK, so, seems that pull_request_target use master as a reference such that PR'ed branch can use secrets (belonging to master) and make docker build OK, but, does not checkout PR branch so the code modified by the PR is actually not tested?!?!... Using pull_request break docker anyway (no access to secrets - docker login/password - in master).

So I'll cut this PR in 2 parts : fix-ci and fix-github-actions. fix-github-actions has a checkout on ${{github.ref}} hoping this makes PR run on PR'ed branch : fix-github-actions will have to be yet another blind-push... No better than this at the time: help is appreciated.

fghoussen · 2022-02-12T11:10:20Z

Now #343 is merged... Maybe this PR will test the code modified by this PR?!... Rebasing on master...

fghoussen · 2022-02-12T11:32:11Z

@sylvestre : #343 didn't help. Do you know how to fix this?

fghoussen · 2022-02-13T12:31:15Z

Check commit is OK

commit a40e66c6c6aa68fac4e97c6aca9992dedb1826c4
Author: Franck HOUSSEN <***@users.noreply.github.com>
Date:   Sun Feb 13 13:22:23 2022 +0100

    run travis_redhat.

But still does not run docker_redhat instead of docker_redhat (which should eb renamed podman_redhat BTW!) ?!...

fghoussen · 2022-02-26T15:43:35Z

redhat still fails?!... The check commit stage points out the correct commit from the correct repo/branch (the PR' ones), but, the job is the one from master?! @sylvestre : ring a bell?

fghoussen · 2022-02-28T18:17:28Z

@sylvestre: PR does not run code from PR'ed branch but the one from master?! PR changes are not tested?!

fghoussen · 2022-03-05T12:38:40Z

Same problem: PR is not tested, ea1dfe7 (master) is ran, not b5c02b0 (PR). No solution. any help is welcome @sylvestre ?

fghoussen · 2022-03-14T11:14:48Z

@sylvestre: my understanding (I may be wrong) is that we must drop all docker/podman (redhat) jobs, switch back to pull_request... Test the new workflow and hope for it to work as expected...

To run docker jobs we now need login/password. These login/password are stored as secrets in opencollab user: any user who would PR can not access them, so all docker builds are KO.
To access them, we need to trigger CI on pull_request_target instead of pull_request: this allows any user who would PR to run in 'the env of opencollab master with secrets as env variables' so docker jobs can run... But it seems that the commit is stuck on opencollab master (ea1dfe7) but not the one the CI is supposed to test (= b5c02b0 = commit of the PR => check out podman_redhat job : the job ran is the one from opencollab master, not the one the PR intend to modify).

Not sure to know how to get out of this mess... Switching back to pull_request will prevent docker to run, and, I'am not even sure this would make things back as expected (= CI testing the code modified by the PR!)

https://securitylab.github.com/research/github-actions-preventing-pwn-requests/

sylvestre · 2022-04-02T12:48:06Z

not sure what you mean?!
many jobs seems to work
just disable the failing jobs?
it is probably good enough

fghoussen · 2022-04-02T13:01:15Z

Less and less time for that...

As is jobs does not work: #341 (comment)

Only solution I see:

drop all docker-like jobs (free the workflow from use of secrets - mandatory for docker login)
return to pull_request (do not use pull_request_target that seems to make the workflow based on master)
test the resulting pull_request workflow (without docker-like jobs and no need for secrets)
make sure the code of a PR is indeed tested by the PR (= PR test the code modified - this is no more what happens!).

This means getting back to CI when done on TravisCI with only distro offered by Travis. Sounds to me like the only way. Not sure if there is a better way

@sylvestre: OK ?

sylvestre · 2022-04-02T13:07:41Z

sounds good!

fghoussen · 2022-04-03T10:27:33Z

OK. If you're OK, I'll try to fix the workflow this way when possible (not much time for now).
The workflow is now broken: when one PR, it seems that the code of master is ran (not the one of the PR)... Looks related to the fact that with github actions, all is done to avoid/limit/prevent malicious actions: so simple things are no more simple...

looks like PR code is not checked out ?!... travis_redhat is not ran : dummmy podman test (from master) is ran instead ?!.

fghoussen · 2022-04-09T16:56:25Z

@sylvestre : drop coverage? Never works with github-actions (but did with TravisCI)

Mac with cmake KO: not mac-based, can somebody help? @prj-: any clue?

[ 99%] Building C object CMakeFiles/icb_parpack_c.dir/PARPACK/TESTS/MPI/icb_parpack_c.c.o
clang: error: unknown argument: '-fallow-argument-mismatch'

prj- · 2022-04-10T06:54:04Z

Where is -fallow-argument-mismatch coming from? It should only be set for the Fortran compiler.

fghoussen · 2022-04-10T13:16:29Z

Where is -fallow-argument-mismatch coming from?

No idea! On github-actions boxes, seems that gcc can not be installed (pkg conflicts): gfortran is used for fortran and clang for C/C++

arpack-ng/.github/workflows/jobs.yml

Line 204 in 8d62927

brew rm gcc

It should only be set for the Fortran compiler.

I know! Not mac-based and not so familiar with it. This is why I tried f36b044. If you have any clue, would be nice :)

fghoussen · 2022-04-10T13:48:42Z

@sylvestre: are you OK with 0ba37a7?

fghoussen · 2022-04-10T16:59:07Z

OK, my understanding is that problems have been passed for C/C++ with gfortran/clang, but, we hit now same problems for MPI (parpack).

[ 99%] Building C object CMakeFiles/icb_parpack_c.dir/PARPACK/TESTS/MPI/icb_parpack_c.c.o
clang: error: unknown argument: '-fallow-argument-mismatch'

fghoussen mentioned this pull request Feb 12, 2022

Support for Microsoft's non-standard complex types #334

Merged

fghoussen force-pushed the fix-ci branch from 5430422 to 8b7c6ea Compare February 12, 2022 11:14

fghoussen force-pushed the fix-ci branch from 8b7c6ea to 36ca23d Compare February 13, 2022 10:25

fghoussen force-pushed the fix-ci branch from 32e72e8 to 8d80d14 Compare February 26, 2022 13:27

fghoussen mentioned this pull request Apr 2, 2022

[c,z]dotc => [cc,zz]dotc #346

Merged

fghoussen added 7 commits April 9, 2022 16:04

mv travis_centos.sh -> travis_redhat.sh.

a60f6cc

redhat: kill sudo that podman doesn't need.

073ca40

redhat: replace docker with podman.

2614a69

redhat: pull redhat image.

60879c5

redhat: ubi as name.

7659cb3

redhat: kill dnf export.

73cfba6

redhat: kill cmake export.

78de283

fghoussen added 7 commits April 9, 2022 16:04

redhat: kill cmd_powertools export.

2ff42bd

redhat: cat release.

ad6798a

run travis_redhat.

47da71e

spy over git before to run podman+redhat :

5796cca

looks like PR code is not checked out ?!... travis_redhat is not ran : dummmy podman test (from master) is ran instead ?!.

Do all by-hand: make things we need, do not use any actions ?...

45ab66c

Kill podman

806dc6d

Kill docker

33bf1b6

fghoussen force-pushed the fix-ci branch from b5c02b0 to 33bf1b6 Compare April 9, 2022 14:19

Fix github actions: checkout branch we need to test - fix mac ?

af24033

fghoussen force-pushed the fix-ci branch from 0a61ff5 to c3c8875 Compare April 9, 2022 16:34

Clang: don’t emit warning for unused driver arguments.

f36b044

fghoussen force-pushed the fix-ci branch from c3c8875 to f36b044 Compare April 9, 2022 16:37

macos - cmake: force compilers ?

e523354

fghoussen added 3 commits April 10, 2022 15:40

same thing but one-lined

3e7dbdb

dummy: CXX=clang++

afdec7a

Kill coverage: never worked with github-actions.

0ba37a7

compiler exports before cmake

999a3ba

macos: force mpi compilers?

e55b4b4

fghoussen merged commit 19a08ec into opencollab:master Apr 10, 2022

fghoussen deleted the fix-ci branch April 10, 2022 17:46

Fix ci #341

Fix ci #341

Uh oh!

Conversation

fghoussen commented Feb 6, 2022

Pull request purpose

Uh oh!

fghoussen commented Feb 7, 2022

Uh oh!

sylvestre commented Feb 7, 2022

Uh oh!

fghoussen commented Feb 9, 2022

Uh oh!

fghoussen commented Feb 10, 2022

Uh oh!

fghoussen commented Feb 10, 2022

Uh oh!

fghoussen commented Feb 10, 2022

Uh oh!

fghoussen commented Feb 12, 2022

Uh oh!

fghoussen commented Feb 12, 2022

Uh oh!

fghoussen commented Feb 12, 2022

Uh oh!

fghoussen commented Feb 13, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

fghoussen commented Feb 26, 2022

Uh oh!

fghoussen commented Feb 28, 2022

Uh oh!

fghoussen commented Mar 5, 2022

Uh oh!

fghoussen commented Mar 14, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sylvestre commented Apr 2, 2022

Uh oh!

fghoussen commented Apr 2, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

sylvestre commented Apr 2, 2022 via email

Uh oh!

fghoussen commented Apr 3, 2022

Uh oh!

fghoussen commented Apr 9, 2022

Uh oh!

prj- commented Apr 10, 2022

Uh oh!

fghoussen commented Apr 10, 2022

Uh oh!

fghoussen commented Apr 10, 2022

Uh oh!

fghoussen commented Apr 10, 2022

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

fghoussen commented Feb 13, 2022 •

edited

Loading

fghoussen commented Mar 14, 2022 •

edited

Loading

fghoussen commented Apr 2, 2022 •

edited

Loading