I would strongly prefer the jll version to match the actual Julia version... otherwise things get confusing.

But what is the "actual julia version" given that this ships 1.6 - 1.10 in various variants :-)

Oh right... Morning brain

Perhaps it should be the maximum of all versions, so here still 1.10 ? Whatever we agree on (or agreeD ON -- I probably just forgot the outcome) I suggest we add it as a comment to avoid repeating the discussion needlessly:-)

Hrm, 1.9 fails in mingw, and 1.10 fails everywhere:

[11:59:22] make[2]: *** No rule to make target '/workspace/srcdir/julia/src/support/hashing.o', needed by '/workspace/srcdir/julia/src/support/libsupport.a'.  Stop.
[11:59:22] make[2]: Leaving directory '/workspace/srcdir/julia/src/support'
[11:59:22] make[1]: *** [Makefile:329: support/libsupport.a] Error 2
[11:59:22] make[1]: Leaving directory '/workspace/srcdir/julia/src'
[11:59:22] make: *** [Makefile:62: julia_flisp.boot.inc.phony] Error 2
[11:59:22] make: *** Waiting for unfinished jobs....

OK, I went one commit back, leaving out Julia commit JuliaLang/julia@41b73e2 ("Bump libuv"). That resolves the 1.10 builds for now.

That leaves the failing mingw builds.

That leaves the failing mingw builds.

Once again it is that msys patch that is not applied since the julia version changed from 1.9.0-DEV to 1.9.0-alpha1 which doesn't match the folder name anymore. Instead of the patch we might also try to remove the fake uname at the beginning of the build:

rm -f /usr/bin/uname

(This reports that it is running in an MSY2 shell instead of a Linux build-os)

But I can't test this right now since for some reason I am hitting that

*** No rule to make target '/workspace/srcdir/julia/src/support/hashing.o', needed by '/workspace/srcdir/julia/src/support/libsupport.a'.

locally, even with the current Yggdrasil master instead of the updates from this branch.

Once again it is that msys patch that is not applied since the julia version changed from 1.9.0-DEV to 1.9.0-alpha1

Urgh, one would think I'd have learned that by now :-(. Thank you, trying that now.

Assuming this works, I'd like to merge and then try the "remove fake uname" trick in a separate PR (it would be nice to not require those patches)