Brad Fitzpatrick

(label "upstreamed" is a bit of a stretch because we didn't do it, but it's the closest label we've got)

> in the extremely unlikely case that someone else is trying to do the same Oh, hi! I'm also working on cloud-ifying (AWS, GCP, Azure, Proxmox, ...) Gokrazy appliance images....

ts1.24: 6fcc25b5d58dbfd0505f1dae0d260d88cf19365e

How's it crash? Got any logs?

What makes you think that? Any logs? What's "adb logcat" say while you attempt to start it?

Ah, thanks. I only saw the other commenter's logs. I wish f-droid would test their builds before publishing them.

@linsui it sure looks like the problem was that prior to you disabling the build in https://gitlab.com/fdroid/fdroiddata/-/commit/00d207cd1981c57fb57607ea6d9d156ecc6886c9#281743176537bad7e7758254b7095f352bea0c57_1554_1555 , the `tailscale-go` srclibs hash was wrong and didn't match the repo's declared...

> the @... is just the clone point, we later checkout the commit based on your source code file How is it guaranteed that the clone point will contain code...

Gotcha. In any case, it seems like the recipe is using a mix of Go toolchains. That `curl` output in there from `./tool/go` (https://github.com/tailscale/tailscale-android/blob/main/tool/go#L13-L71) means that `TOOLCHAINDIR` is not set,...

@linsui, the `curl` is coming via the `TAILSCALE_VERSION=$(shell ./version/tailscale-version.sh 200)` in the Makefile, which sets `TS_USE_TOOLCHAIN=1` (it probably shouldn't) which then calls into the `tailscale/tailscale` repo's `build_dish.sh`, which does: ```sh...