is there any way to introduce Advanced Student Language Signatures https://docs.racket-lang.org/htdp-langs/advanced.html#%28part._advanced-signatures%29 , much earlier ? If not, I think I might try to jump to ASL right after BSL. Ahh, found the reason why they are disabled: https://github.com/racket/racket/issues/2941#issuecomment-559304146

@popa.bogdanp it looks like the default setup-racket that comes with raco pkg new will fail currently. See e.g. https://github.com/sorawee/pprint-compact/runs/3862069047?check_suite_focus=true. Do you know what’s going on?

I guess a part of it is that HEAD is now 8.3, but it doesn’t quite work yet.

IIRC that is a pretty ancient version of setup-racket.

Should be updated then!

It still fails even with the latest version

https://github.com/racket/racket/blob/master/racket/collects/pkg/private/new.rkt#L144

I was seeing a failure this morning but it had to do with building on current

Mine is weird. currents fail with:

Unpacking into "/usr/racket-8.3.0.1-tmp-install" (Ctrl+C to abort)... Done. /usr/racket-8.3.0.1-tmp-install/bin/racket: error while loading shared libraries: libffi.so.6: cannot open shared object file: No such file or directory Error: installation failed and stables are cancelled.

So none succeeds

Yes that’s the same error I’m getting

OK, so the issue is workaround by removing current from the matrix. This allows stable to finish without being cancelled.

Also the snapshot (“current”) builds are 404-ing starting a couple days ago: Run Bogdanp/setup-racket@v1.5 Installing Racket current (BC, full, x64)... Error: Unexpected HTTP response: 404 But that’s not an issue for @popa.bogdanp. In my experience, if there is a holiday, then there is a non-zero chance a university IT system will be down for most of it, intentionally or otherwise. :P

Did the current snapshot package stop having libffi in it?

Not sure if this is a @popa.bogdanp bug or a @mflatt bug

@sorawee As for “stable”, it looks yours is BC, but the CS flavor is working for me, at least as of a few minutes ago: https://github.com/greghendershott/racket-mode/runs/3862550812?check_suite_focus=true

Right. I think the problem is a cascading failure from current which somehow makes stable “operation cancelled”. After removing current from the matrix, stable succeeds

Mine has both “current” and “stable” (w/o specifying a flavor, it gets CS) and “stable” works there. FWIW

I haven’t been noticing the 404 issue. But it might be good to let @mflatt or @samth know

(I mean “current” does fail for me with a 404, and good thinking to @ them. But its failure doesn’t seem to affect “stable” (default, CS) for me.)

(Also not saying you’re wrong to test against BC. Just sharing data points.)

I noticed the failures w/ the latest BC snapshots this morning, but haven’t had time to dig in. I think it’s a bug in the BC snapshot.

I just tried installing https://www.cs.utah.edu/plt/snapshots/current/installers/racket-current-x86_64-linux-precise-bc.sh\|it in a Ubuntu 20.04 container and it fails with the same libffi error.

> I think the problem is a cascading failure from current which somehow makes stable “operation cancelled”. After removing current from the matrix, stable succeeds The default behavior on GHA is to cancel all jobs when the first one fails. You can add fail-fast: false under strategy (as a sibling of matrix) in your config to change that behavior (https://github.com/Bogdanp/setup-racket/blob/379849dbba606cf4f790738cd33125b959f9632d/.github/workflows/workflow.yml#L13\|example).

Aha! Thanks @popa.bogdanp

I moved the Utah snapshots from Ubuntu 12.04 to Debian 8, so it’s probably related to that. Maybe the new build installation doesn’t have the right thing, or maybe too much. I can investigate, but not until much later today.

The problem was that the Debian 8 installation included libffi, so the Racket build linked to it instead of building its own. I’ve adjusted the configuration to fix that, so the next snapshot should be ok.

It looks like “setup-racket” includes “precise” in the name use for Linux installers from the Utah site. That’s now “jesse”, although I set up “precise” aliases to keep them (and I guess that much worked). But there’s also an alias that omits the distribution name, and the OS-less name really the intended name to use for this purpose. The download page doesn’t currently make various aliases discoverable enough.

Thanks! I’ll update setup-racket to drop the distro bits from the URLs in that case. It looks like the pre-release build of CS for Windows (http://pre-release.racket-lang.org/installers/racket-current-x86_64-win32-cs.exe\|url) is also failing:

https://github.com/Bogdanp/setup-racket/pull/45/checks?check_run_id=3866682917

When I try it in a Windows 10 VM, I get an error that the installer integrity check failed.