I tried to verify the fix by doing a basic tmt run via container.

Then I moved into a blank directory with no fmf tree and tried to run with the recipe file:
(tmt version: 1.71.0)

tcornell ~/tmp $ tmt run -vvv --recipe /var/tmp/tmt/run-063/recipe.yaml

No metadata found in the '.' directory. Use 'tmt init' to get started.

The exception was caused by 1 earlier exceptions

Cause number 1:

    Unable to find tree root for '/home/tcornell/tmp'.

(PR)

(dev) tcornell ~/tmp $  tmt run -vvv --recipe /var/tmp/tmt/run-063/recipe.yaml

[...]

    execute
        queued execute tasks
            #1: default-0 on default-0
        
        execute task #1: default-0 on default-0
        how: tmt
        order: 50
        ignore-duration: false
        exit-first: false
            test: Concise summary describing what the test does
                duration deadline: 300 seconds, at 16:23:03 2026-04-20 UTC
                cmd: ./test.sh
                stdout: ./tmt-test-wrapper-inner.sh-smoke-1-default-0-default-0: line 1: ./test.sh: No such file or directory
                00:00:00 errr /smoke (on default-0) [1/1]
                    Note: beakerlib: TestResults FileError

    
        summary: 1 test executed

Looks like it made the fake tree, but the tree does not have the supporting test files.
When I execute the same command in the original fmf root directory, it runs fine.

Am I reproducing this correctly?

Looks like it made the fake tree, but the tree does not have the supporting test files. When I execute the same command in the original fmf root directory, it runs fine.

Am I reproducing this correctly?

Yes, I would say that is expected. When users move the recipe or execute it from a different location, they should handle the necessary test files. Either move them to the new directory or rewrite the recipe with updated paths.

I think I am misunderstanding the use case. What situations would a user execute the tmt command outside of the fmf tree? Could you provide an example or add a test case, please?

I think I am misunderstanding the use case. What situations would a user execute the tmt command outside of the fmf tree? Could you provide an example or add a test case, please?

In some cases, an fmf tree is not needed when the run is loaded from the recipe, unless the test requires some additional files. I ran into this while creating a script for re-executing tests. tmt would raise an exception due to a missing tree, even though we wouldn't need it.

But looking at it again, I completely missed that we actually save the fmf root path in the recipe, so we can just set it from there, and that would solve the issue. Changed in f2512fc.

But looking at it again, I completely missed that we actually save the fmf root path in the recipe, so we can just set it from there, and that would solve the issue. Changed in f2512fc.

👍 to the idea, but can you double check that the tree in there is the user tree and not the work tree. If it is the latter, then it should be changed to the former imo.

What situations would a user execute the tmt command outside of the fmf tree?

Not exactly the original context, but a situation that should probably be supported (or disallowed explicitly) is if one runs tmt run ... without a fmf root at all.

👍 to the idea, but can you double check that the tree in there is the user tree and not the work tree. If it is the latter, then it should be changed to the former imo.

It is a user tree, e.g. /home/user/tmt-plans

freedesktop.org is down, hence the linkcheck fail. The rest is green, merging.