eval: regression in use of close with matchN

[myitcv]

What version of CUE are you using (cue version)?

$ cue version
cue version v0.0.0-20250122112448-ac4c7d0c2f80

go version go1.23.2
      -buildmode exe
       -compiler gc
  DefaultGODEBUG asynctimerchan=1,gotypesalias=0,httpservecontentkeepheaders=1,tls3des=1,tlskyber=0,x509keypairleaf=0,x509negativeserial=1
     CGO_ENABLED 1
          GOARCH arm64
            GOOS linux
         GOARM64 v8.0
             vcs git
    vcs.revision ac4c7d0c2f80d276a134f6e0eebd3b4c724854eb
        vcs.time 2025-01-22T11:24:48Z
    vcs.modified false
cue.lang.version v0.12.0

Does this issue reproduce with the latest release?

n/a - working with tip.

What did you do?

env CUE_EXPERIMENT=evalv3

exec cue eval x.cue

-- x.cue --
#step: matchN(1, [{
	"uses"!: _
	...
}, {
	"run"!: _
	...
}]) & close({
	"uses"?: string
	"run"?:  string
})

What did you expect to see?

Passing test.

What did you see instead?

> env CUE_EXPERIMENT=evalv3
> exec go run cuelang.org/go/cmd/cue eval x.cue
[stderr]
#step: invalid value {uses?:string,run?:string} (does not satisfy matchN): 0 matched, expected 1:
    ./x.cue:1:8
    ./x.cue:1:15
exit status 1
[exit status 1]
FAIL: /tmp/testscript198252630/repro.txtar/script.txtar:6: unexpected command failure

This bisects to aa6781f, and is currently a blocker for curated modules.

[mpvl]

It is correct it is an error, but it should be an incomplete error, instead of terminal error. (i.e. eval should indeed pass.)

[myitcv]

@mvdan is absolutely correct in that I should have reported the root problem here, which I do below, using cue export and not cue eval:

What did you do?

env CUE_EXPERIMENT=evalv3

exec cue export x.cue
cmp stdout stdout.golden

-- x.cue --
#step: matchN(1, [{
	"uses"!: _
	...
}, {
	"run"!: _
	...
}]) & close({
	"uses"?: string
	"run"?:  string
})

s: #step & {
	run: "echo hello world"
}
-- stdout.golden --
{
    "s": {
        "run": "echo hello world"
    }
}

What did you expect?

Passing test.

What did you see instead?

> env CUE_EXPERIMENT=evalv3
> exec cue export x.cue
[stderr]
#step: invalid value {uses?:string,run?:string} (does not satisfy matchN): 0 matched, expected 1:
    ./x.cue:1:8
    ./x.cue:1:15
[exit status 1]
FAIL: /tmp/testscript364055009/repro.txtar/script.txtar:3: unexpected command failure

The analysis/conclusion etc might be exactly the same, but either way for completeness' sake I capture the real trigger for this in our use case.

[mpvl]

It was more intended as an analysis and note to self. :)

Found the cause, but not the solution. The bisect CL is correct, but exposes this preexisting bug.