pocmatos
2020-4-22 08:40:14

Was thinking about the script to build/test racket, which I would prefer to write in Racket but…


pocmatos
2020-4-22 08:40:33

If the idea is to use them to build racket cross-arch, this means we need a racket cross-arch.


pocmatos
2020-4-22 08:41:40

hummm, actually if we use a debian cross-arch container, debian ships racket at least for arm so we could use that.


pocmatos
2020-4-22 08:42:18

Otherwise for other archs, we could create an unofficial distribution just to CI racket ports.


pocmatos
2020-4-22 08:43:37

ok, thanks.


samth
2020-4-22 12:53:36

What if we put the prefix somewhere known, like $HOME/install, and create that directory?


samth
2020-4-22 12:53:47

Then it doesn’t matter what GitHub does


pocmatos
2020-4-22 14:32:17

That would work but the problem is that I don’t know what kind of permissions we have. We can try… :slightly_smiling_face:


samth
2020-4-22 14:32:30

we can run as root if needed


samth
2020-4-22 14:32:46

also we should definitely have permissions under $HOME


pocmatos
2020-4-22 14:32:52

I am not even sure which user we run as… maybe I have seen it before but cannot remember.


pocmatos
2020-4-22 14:33:04

Will create a dummy project and test.


pocmatos
2020-4-22 14:39:07

@samth no point in talking in thread when it’s just us. :slightly_smiling_face:



pocmatos
2020-4-22 14:39:18

You are right!


pocmatos
2020-4-22 14:39:35

We are user runner , home is /home/runner


pocmatos
2020-4-22 14:39:40

and we have permissions at home.


pocmatos
2020-4-22 14:39:52

Lets install everything to $HOME/install?


samth
2020-4-22 14:44:00

yes, that sounds good to me


samth
2020-4-22 14:44:12

can you do that? I’m not sure I’d get all the places.


pocmatos
2020-4-22 14:45:27

I can do then - not sure I can reuse your PR. Might have to create a new one.


pocmatos
2020-4-22 14:45:35

If that’s ok with you, I will do it now.


samth
2020-4-22 14:46:45

Yes, I can just close mine.


pocmatos
2020-4-22 14:50:20

OK - I was looking at the docs because we cannot use $HOME in working-directory: tags.


pocmatos
2020-4-22 14:50:49

It turns out that github has ${{ runner.temp }} to use which might be a good idea to use. Docs here: https://help.github.com/en/actions/reference/context-and-expression-syntax-for-github-actions#runner-context


pocmatos
2020-4-22 14:50:57

What do you think?


samth
2020-4-22 14:53:46

that also seems fine


samth
2020-4-22 14:53:57

but why do we need to set working-directory?


pocmatos
2020-4-22 15:04:16

For upload/download artifact actions. You don’t have access to the shell to change directory.


pocmatos
2020-4-22 15:04:51

pocmatos
2020-4-22 15:05:07

will be impressed if I got all the paths right. Need to leave now for a bit, will come back later.


pocmatos
2020-4-22 15:05:18

Lets see what CI says in the meantime.


samth
2020-4-22 15:54:31

There are still several uses of .. in the file — in the windows build and in the scanbuild steps. Should those change too?


pocmatos
2020-4-22 16:45:09

right - there’s a missing one, now fixed. All good with .. in scan-build report mgm and windows build


samth
2020-4-22 16:50:41

Do we need to change the ci-pr file?


pocmatos
2020-4-22 18:43:30

Don’t think so… there was a problem… with working directory on a uses. Will fix.


pocmatos
2020-4-22 20:12:03

Wow - I am stumped by the current failures in scanbuild jobs for 3m and cs: https://github.com/racket/racket/runs/609803669?check_suite_focus=true#step:8:207


pocmatos
2020-4-22 20:12:27

They can’t find racketcgc binary but configure just said in the previous step the binary path is correct.


samth
2020-4-22 20:17:40

it seems to be building 3m in the previous step, and then not finding cgc?


pocmatos
2020-4-22 20:19:29

right…


pocmatos
2020-4-22 20:19:47

although I just noticed that when configure says: hey, we are using racket FOO, it doesn’t actually check FOO exists.



pocmatos
2020-4-22 20:22:11

It’s getting late here. I would like to fix the current scanbuild issue but unsure if I will make it.