dhewg

D'uh, that works too, done

If that works and is acceptable you can move the file and function to `target/linux/ath79/image/`. This isn't the first of its kind too: `target/linux/lantiq/image/eva.dummy.squashfs`. Burried in 0f6a9d5c7c938f40657ca30f588479893606781e

Isn't that what PKG_SOURCE_URL_FILE is for?

Alright. But why not simply use `PKG_SOURCE:=squashfs$(PKG_VERSION).tar.gz` then? That's an official release, why muck around with it? Having 3.0 and 4.2 in the dl folder is pretty clear. Is there...

That was my first idea, but then there's netgear: https://github.com/openwrt/openwrt/pull/10967#issuecomment-1419327294

> fwiw, https://git.alpinelinux.org/aports/tree/main/zstd/APKBUILD#n55 This host package is already build with meson

The same commit switched from make to meson too, and the comment mentions Makefile, so I assume that refers to that part? Maybe not, maybe both ;)

Sure, I can do that. To me a revert isn't strictly a problem, it's just a reverse patch for whatever reason

Huh, I think I never did it that way around. I can try to test that. Assuming that worked before and with these patches not anymore, do we want to...

Without this PR, diff of `find . -type f|sort|xargs sha256sum` between fresh full toolchain build, then wiped staging_dir/toolchain, followed by a toolchain/compile: ``` --- before.sha256sums 2023-02-09 21:24:28.143571670 +0100 +++ after.sha256sums...