Can't use computed lists in matrix loops #2544
Description
Description
You can only use literal lists in matrix loops, either by reference, or directly. Using a list created from some other source, fails.
See the example taskfile, where OS_KEYS is a list produced by the keys template command.
Output for task:
[spew] OS_VAR: ([]interface {}) (len=3 cap=3) { [spew] (string) (len=7) "windows", [spew] (string) (len=5) "linux", [spew] (string) (len=6) "darwin" [spew] } [spew] OS_KEYS: ([]string) (len=3 cap=4) { [spew] (string) (len=6) "darwin", [spew] (string) (len=7) "windows", [spew] (string) (len=5) "linux" [spew] } [list-ref] windows/amd64 [list-ref] windows/arm64 [list-ref] linux/amd64 [list-ref] linux/arm64 [list-ref] darwin/amd64 [list-ref] darwin/arm64 task: Failed to run task "default": task: Failed to parse /tmp/repro/Taskfile.yml: matrix reference ".OS_KEYS" must resolve to a list I do note that spew shows that the keys list a []string array, while OS_VAR is a []interface {} array, but I wouldn't expect that to be the issue.
Version
3.45.5
Operating system
macos 15.6.1
Experiments Enabled
No response
Example Taskfile
version: '3' output: prefixed vars: OS_MAP: map: windows: "Microsoft Windows" linux: "Linux kernel" darwin: "Apple macos" OS_VAR: ["windows", "linux", "darwin"] OS_KEYS: { ref: keys .OS_MAP } ARCH_VAR: ["amd64", "arm64"] tasks: spew: silent: true cmds: - "echo -n 'OS_VAR: {{spew .OS_VAR}}'" - "echo -n 'OS_KEYS: {{spew .OS_KEYS}}'" list-ref: silent: true cmds: - for: matrix: OS: { ref: .OS_VAR } ARCH: { ref: .ARCH_VAR } cmd: echo "{{.ITEM.OS}}/{{.ITEM.ARCH}}" keys-output: silent: true cmds: - for: matrix: OS: { ref: .OS_KEYS } ARCH: { ref: .ARCH_VAR } cmd: echo "{{.ITEM.OS}}/{{.ITEM.ARCH}}" default: cmds: - task: spew - task: list-ref - task: keys-output