I am seeing this a lot:

https://github.com/racket/racket/runs/2956096253#step:4:8

even in my own repos.

any idea why?

http://status.github.com\|status.github.com says everything is fine… sigh

Does setup-racket use http://mirror.racket-lang.org\|mirror.racket-lang.org (which is down) or http://download.racket-lang.org\|download.racket-lang.org (which should redirect to the current mirror)?

It uses http://mirror.racket-lang.org\|mirror.racket-lang.org

oh - hummm, that could be it.

:disappointed:

Racket News issue 52 is stuck trying to publish because of that.

I think the reason for that is that if you go to http://download.racket-lang.org\|download.racket-lang.org, it hyperlinks to http://mirror.racket-lang.org\|mirror.racket-lang.org (normally, right now it hyperlinks to http://cs.utah.edu\|cs.utah.edu)

So it’s not easy to discover that you can or should link to http://download.racket-lang.org\|download.racket-lang.org

Yes, I agree that it makes sense to change that.

I’ll try a little later this morning.

Should this: https://github.com/Bogdanp/setup-racket/blob/c224e7297b5b54745f28b5e74d99f68eb3fc73db/src/common.ts#L50

be http://download.racket-lang.org\|download.racket-lang.org ?

as in let base = <https://download.racket-lang.org/installers/${version}>;

?

@popa.bogdanp ^^^^

yes, I think that’s the right change

I’ve changed the download web pages, and things seem to work ok there.

They work for me too.

does that mean setup-racket can keep using mirror?

no

it means people in the future will be less likely to think they should use mirror

mirror is a machine at northeastern that currently is turned off

https://github.com/Bogdanp/setup-racket/pull/37

ok lets hope @popa.bogdanp merges this then.

Seems like some tests are failing with the new URL: https://github.com/Bogdanp/setup-racket/runs/2964058645

That one tried to download https://download.racket-lang.org/installers/8.1/racket-8.1-x86_64-macosx.dmg

Looks like the failures are for BC variants so I assume the Utah site needs the -bc suffix for versions 8.0 and up where the other one didn’t

Oh no - do you have time to get a fix sorted for that?

Yeah, I’m working on that now

I guess the Utah mirror doesn’t have links for the default variants.

Leaving out “-cs” or “-bc” would get “-cs” in the case of v8.0 and v8.1.

Thanks @popa.bogdanp

Yeah, I think that was a bug in setup-racket anyway. Anyone asking for BC 8.0 would’ve received CS 8.0, b/c it didn’t add the suffix

@pocmatos can you give v1.5 a spin? https://github.com/Bogdanp/setup-racket/releases/tag/v1.5

running - lets see if we get a draft version of RN52 published. :slightly_smiling_face:

https://github.com/pmatos/racket-news/actions/runs/990655272

You may need to re-run it once. I forgot to publish the built artifacts initially.

yeah that one seems to still be using mirror https://github.com/pmatos/racket-news/runs/2964334777?check_suite_focus=true#step:5:11

Yeah, I realized I’d forgot to publish the artifacts just as I typed out my other message :sweat_smile:

Looks like it works!

I’d recommend using the full distribution (distribution: full in YAML), btw. It would probably speed up your builds significantly based on the dependencies you have

Thanks, will do.

Excelent - thanks @popa.bogdanp - also update the racket repo to use v1.5

https://github.com/racket/racket/commit/8c140a85eb31a61c8d70112177dbdbdae7c7fcaa

@popa.bogdanp i wonder what changed… now using v1.5, building docs fails with : “raco setup: docs failure: setup: install SQLite to build documentation”

Yeah, I was surprised by that failure. I wonder if the Northeastern builds (accidentally?) shipped SQLite .so s

IIRC, there’s no package to distribute sqlite on Linux in the standard distribution

@mflatt any idea?

sqlite3 normally needs to be installed at the Linux package-manager level to work with docs; but that’s not something new

yeah, this was not failing when installing CS8.0 from mirror.

but now from download, it is failing.

the only thing that changed on the racket news side was the upgrade from setup-racket v0.10 to v1.5

Is there a raco setup that wasn’t happening with older versions of setup-racket, or something like that?

No, the only difference is the installer that’s used

but before setup racket was not trying to build documentation.

no docs: https://github.com/pmatos/racket-news/runs/2842453311?check_suite_focus=true#step:5:1250

@pocmatos From the GitHub Actions log, it looks like you switched from minimal to full. Is that not right?

docs: https://github.com/pmatos/racket-news/runs/2964397519?check_suite_focus=true#step:5:248

@mflatt that’s the last commit. if fails even before that commit.

the commit : “Update setup-racket to v1.5” still uses minimal and fails with sqlite problem as well.

The action just runs the installer with echo "yes\n1\n" \| sh ${path} --create-dir --unix-style --dest /usr/ so it doesn’t really control what gets built. The installer decides.

i mean v0.10 was quite old… maybe something changed in between then and v1.5…

@popa.bogdanp how are the packages installed?

that installs racket but then the failures is while installing packages.

i see that setup-racket changed the way it presents the installation in the github actions log

Ah, I didn’t realize you were using the packages syntax

Yeah, the --no-docs flag was dropped at some point between 0.10 and now so that’s it

where you purposefully passing that flag and you dropped it?

Yes, but I’d have to look up why I did that

no worries… is there a way to pass custom flags to setup?

No. I’d just change your worflow to drop the packages entry and add a job like - run: raco pkg install ... with all the right flags

ok - thanks.

thanks - lets wait and see :slightly_smiling_face:

https://github.com/Bogdanp/setup-racket/issues/18#issuecomment-755144715 was the reasoning behind that

I think --no-docs is dropped in package: because earlier Racket 7.x does not support it