`forced-target` does not work with `host-tuple`

[thealtofwar]

Problem

When using forced-target on a package, host-tuple isn't substituted properly and causes an issue

[package]
name = "pkg"
version = "0.1.0"
edition = "2024"
forced-target = "host-tuple"

[dependencies]

Steps

Check https://github.com/thealtofwar/forced-target-host-tuple-repro

Possible Solution(s)

Apply the same host-tuple change that is applied to --target to this

Notes

No response

Version

cargo 1.93.0-nightly (2d4fa1395 2025-11-12)
release: 1.93.0-nightly
commit-hash: 2d4fa139552ebdd5f091a1401ed03f7dc62cb43f
commit-date: 2025-11-12
host: x86_64-pc-windows-msvc
libgit2: 1.9.1 (sys:0.20.2 vendored)
libcurl: 8.15.0-DEV (sys:0.4.83+curl-8.15.0 vendored ssl:Schannel)
os: Windows 10.0.26200 (Windows 11 Professional) [64-bit]

[weihanglo]

Should probably try to make it work, if not too complicated. However, I wonder what the use case is. Mind sharing a bit more how it is found and why you need it?

[thealtofwar]

Should probably try to make it work, if not too complicated. However, I wonder what the use case is. Mind sharing a bit more how it is found and why you need it?

My use case is that I'm writing a kernel, which needs a specific host triple x86_64-unknown-none to function. The kernel is split into two parts, with the device controls needing the target x86_64-unknown-none in one package, and the network packet parsers in another package.

My network packet parsers have their own tests, but these tests won't run if compiled for x86_64-unknown-none, so I want to force these to run as the host so cargo test functions.

The code for this project is available at https://github.com/thealtofwar/conc_os with the current workaround of doing forced-target = "<my host triple>". Perhaps I'm doing something silly and could avoid this issue, please tell me if this is so.

[epage]

This is an interesting case; thank you for sharing the use case!

As some quick context, I overall feel more positive towards a required-target field (skip if not matching) than a forced-target field, see rust-lang/rfcs#3759. That would support target-tuple's or cfgs.

My question is what semantics are trying to be conveyed by `forced-target = "" that we would want to express.

Is a host-tuple being used to represent

• a platform-tuple with std
• a platform-tuple with host tools
• something else?

Understanding this can help us better understand what a solution would look like for rust-lang/rfcs#3759, independent of any shot-term solutions we may do for forced-target. For example, is host-tuple right or adding some kind of cfg that can then be composed with other expressions?