runtime: Go 1.25 panics when emulating linux/amd64 on macOS

[mdempsky]

Go version

go version go1.25.1 linux/amd64

Output of go env in your module/workspace:

runtime: taggedPointerPack invalid packing: ptr=0xffff35774e00 tag=0x1 packed=0xffff35774e000001 -> ptr=0xffffffff35774e00 tag=0x1
fatal error: taggedPointerPack

goroutine 1 gp=0xc000002380 m=0 mp=0x11295c0 [running, locked to thread]:
runtime.throw({0xbbf389?, 0x0?})
	runtime/panic.go:1094 +0x48 fp=0xc000114890 sp=0xc000114860 pc=0x47d388
runtime.taggedPointerPack(0xffff35774e00, 0x1)
	runtime/tagptr_64bit.go:60 +0x12a fp=0xc0001148c8 sp=0xc000114890 pc=0x46630a
runtime.netpollopen(0x1149b60?, 0x11499a8?)
	runtime/netpoll_epoll.go:52 +0x3e fp=0xc000114930 sp=0xc0001148c8 pc=0x44085e
[... elided ...]

What did you do?

When I run podman run -it --platform=linux/amd64 golang:1.25-alpine go version on either of my Mac machines, it immediately fatal errors in taggedPointerPack.

If I run podman run -it --platform=linux/amd64 golang:1.24-alpine go version it works fine.

One of my Macs is an M3 MacBook Air running macOS 26.1 Beta. The other is an M4 MacBook Pro running macOS 26.0.1.

/cc @randall77

What did you see happen?

Runtime fatal error.

What did you expect to see?

"go version go1.25.1 linux/amd64"

[randall77]

Dup of #69255?

[mdempsky]

Dup of #69255?

Possibly. But I'm actually running into that issue trying to build from tip to verify. :/

I'll try bootstrapping from a genuine Intel machine in the morning. Thanks.

[gabyhelp]

Related Issues

• runtime: AIX build broken with fatal error 'taggedPointerPack invalid packing' #73515 (closed)
• go/build: Segfault on Go 1.6.2 darwin/amd64 #17912 (closed)
• runtime: Unable to build go project on macOS 10.12 beta 4 #16575 (closed)
• runtime: panic fatal error: malloc deadlock #24059 (closed)
• runtime: Panic in golang compiler on macOS Sierra Third Public Beta (16A270f) #16585 (closed)
• cmd/compile: panic running code that uses generic interface value #51303 (closed)
• runtime: runtime crashes, reproducer same as #20427 #36371 (closed)
• runtime: After update to Mac OSX, GO BUILD ending up in fatal error: unexpected signal during runtime execution [signal 0xb code=0x1 addr=0x1e069c6ce02 pc=0xf47b] #17367 (closed)
• container/list: panic: runtime error: invalid memory address or nil pointer dereference #53351

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[mdempsky]

I'm still trying to root cause this.

Using Apple's container tool, I checked out Go 1.25.1 and all.bash mostly worked. Some crypto tests failed, which I suspect could be imprecise floating-point emulation affecting the optimized implementations, but haven't investigated yet. It also hung after finishing testing std and printing "FAIL".

Not sure why Podman Desktop is having trouble with Go 1.25.1 when allegedly it's also using Apple's Virtualization.framework and Rosetta.

I plan to try Rancher Desktop and other implementations for comparison too to narrow down where the failure is happening.

[mdempsky]

Using OrbStack with Rosetta enabled, the crypto tests pass, and go tool dist no longer hangs after go test std fails. (The test suite still fails for other reasons, but I think unrelated reasons; e.g., alpine's "git" package seems to exclude the "http-backend" tool.)

[Jorropo]

bisect ?

[mdempsky]

Dup of #69255?

Yeah, that seems to have been it. I tested the two nearest golang:tip-alpine images to commit 2a7f1d4:

• before: podman run --platform=linux/amd64 --rm -it golang:tip-20250831-alpine go version panics.

• after: podman run --platform=linux/amd64 --rm -it golang:tip-20250907-alpine go test regexp passes.

The patch seems non-intrusive, so I'd welcome a back port to Go 1.25.x. But otherwise, I'll just look into running my container builds on a native amd64 host.

Thanks!

[randall77]

I'm not sure this qualifies for a backport. amd64 simulation on arm64 is not an official platform. And arguably, this is a bug in the simulator.