Fork me on GitHub
#babashka-sci-dev
<
2023-09-24
>
borkdude12:09:29

libsci tests broke on windows, randomly starting yesterday: https://github.com/babashka/sci/actions/runs/6290003499 if anyone wants to have a look, go ahead

lispyclouds12:09:02

feels like some weird code gen from a Rust proc macro which is tripping the compiler. seems to have made an enum called: enum_(unnamed_at_D_\\a\\sci\\sci\\libsci\\target\\graal_isolate_h_35_1)

borkdude12:09:47

but why suddenly this change, I mean there is a rust cargo.lock file right

lread15:09:07

Yeah probably a change to Windows image? If I look at the last successful run, I see: > Runner Image > Image: windows-2022 > Version: 20230820.1.0 > Included Software: https://github.com/actions/runner-images/blob/win22/20230820.1/images/win/Windows2022-Readme.md > Image Release: https://github.com/actions/runner-images/releases/tag/win22%2F20230820.1 And if I Iook at the failing run, I see: > Runner Image > Image: windows-2022 > Version: 20230918.1.0 > Included Software: https://github.com/actions/runner-images/blob/win22/20230918.1/images/win/Windows2022-Readme.md > Image Release: https://github.com/actions/runner-images/releases/tag/win22%2F20230918.1 If you look at included software links, you'll see that Rust has been bumped, from failing: https://github.com/actions/runner-images/blob/win22/20230918.1/images/win/Windows2022-Readme.md#rust-tools • Cargo 1.72.0 • Rust 1.72.0 • Rustdoc 1.72.0 • Rustup 1.26.0 from working: https://github.com/actions/runner-images/blob/win22/20230820.1/images/win/Windows2022-Readme.md#rust-tools • Cargo 1.71.1 • Rust 1.71.1 • Rustdoc 1.71.1 • Rustup 1.26.0 So maybe related to that? Dunno.

lread15:09:10

Gotta go water my neighbour's garden, tata for now!

lread17:09:29

Hmm... ubuntu is still using 1.71.1 et al but macOS is using 1.72.0 et al (and passes) so maybe a c compiler change or something.

lread17:09:15

Deciphering versions... seems like... was using Visual Studio 1.7.1 now using Visual Studio 1.7.4...

lread18:09:53

If nobody else gets to it, I'll try to replicate on my Windows VM later today.

lread18:09:30

But now.... I must go for a walk and enjoy the big blue room.

lread00:09:54

Blech, upgrading my Windows VM to VS2022 is not going great.

lread00:09:35

Oh crikey. Word to the Windows-naive: don't blindly run GraalVM install instructions. If you take this advice but run it from Powershell:

setx /M PATH "C:\Progra~1\Java\<graalvm>\bin;%PATH%"
The %PATH% part is not recognized by Powershell, so you end up blowing away your previous value with the constant %PATH%.

lread03:09:07

Ok, I updated my Windows VM. I also installed llvm via winget, that seems to be need for bindgen? I needed gcc, it seems, so I installed mingw via scoop. Was that the right thing to do? But now I'm getting different errors than above when trying to compile Rust:

Compiling Rust!
   Compiling from-rust v0.1.0 (Z:\babashka\sci\libsci\from-rust)
error[E0412]: cannot find type `graal_isolate_t` in this scope
  --> src\
   |
22 |         let mut isolate: *mut graal_isolate_t = ptr::null_mut();
   |                               ^^^^^^^^^^^^^^^ not found in this scope

error[E0412]: cannot find type `graal_isolatethread_t` in this scope
  --> src\
   |
23 |         let mut thread: *mut graal_isolatethread_t = ptr::null_mut();
   |                              ^^^^^^^^^^^^^^^^^^^^^ not found in this scope

error[E0425]: cannot find function `graal_create_isolate` in this scope
  --> src\
   |
25 |         graal_create_isolate(ptr::null_mut(), &mut isolate, &mut thread);
   |         ^^^^^^^^^^^^^^^^^^^^ not found in this scope

error[E0425]: cannot find function `graal_tear_down_isolate` in this scope
  --> src\
   |
33 |         graal_tear_down_isolate(thread);
   |         ^^^^^^^^^^^^^^^^^^^^^^^ not found in this scope
I don't know Rust enough yet to diagnose what's up.

borkdude09:09:54

Thanks for all the findings so far, let's post summary in https://github.com/babashka/sci/issues/902 and continue there

borkdude19:09:03

I think I'm going to disable the tests for now, perhaps the issue will fix itself at some point, no time to dig into this soon

borkdude19:09:17

but if someone has a fix, I'll gladly take it

lread21:09:57

I might look again, but I'll also likely get distracted before I do....