cwl-runner validate fails after upgrade

[BJWiley233]

The files are all existing and in correct folder structure after I moved them. Now when I run cwl-runner --validate I get this error below and org.w3id.cwl.cwl.File is no where in this yml file.

Tool definition file:///Users/****/****/UKBB/analysis-workflows/definitions_v1.1/tools/vep.cwl failed validation:
definitions_v1.1/tools/vep.cwl:5:1:                       checking field `requirements`
...more lines
...more lines
definitions_v1.1/types/vep_custom_annotation.yml:20:11:                   Field `type` references
                                                                          unknown identifier `org.w3id.cwl.cwl.File`,
                                                                          tried
                                                                          file:///Users/brian/Bolton/UKBB/analysis-workflows/definitions_v1.1/types/vep_custom_annotation.yml#vep_custom_annotation/annotation/org.w3id.cwl.cwl.File,

The line is under type: File below

  annotation:
    type:
      type: record
      name: info
      fields:
        file:
          type: File
          label: 'File to be used for annotation, include index file'

However when I run cwl-runner validate on the vep.cwl tool itself it works

[tetron]

So if I understand correctly, before upgrading, it validates correctly, but the upgraded version fails to validate?

There were some recent changes in the behavior of schema-salad, and cwl-upgrader has an unpinned dependency on schema salad, so that seems related.

Is this part of a SchemaDefRequirement?

[BJWiley233]

Yes it is! under vep.cwl tool:

requirements:
- class: InlineJavascriptRequirement
- class: SchemaDefRequirement
  types:
  - $import: ../types/vep_custom_annotation.yml

I already have schema-salad installed

[mr-c]

@BJWiley233 Can you try with https://pypi.org/project/schema-salad/8.0.20210624154013/ instead of the latest version?

[BJWiley233]

Ah you replied at exact same time I updated. Let me check.

Anyways I got same error. Any time I have a type: File it fails. Examples are these 2 types:

https://github.com/genome/analysis-workflows/blob/master/definitions/types/vep_custom_annotation.yml
https://github.com/genome/analysis-workflows/blob/master/definitions/types/trimming_options.yml

I get same error with schema-salad-tool vep_custom_annotation.yml along lines of:

../../definitions/types/vep_custom_annotation.yml:19:19:         Field `type` references unknown
                                                                 identifier `File`, tried

[BJWiley233]

Actually I just tested and my original v1.0 again with cwl-runner --validate and it doesn't work. My cwl-runner must have been updated.

[tetron]

Ok, so that is a regression in cwltool. The workaround is to go back a version or two.

Is your CWL publicly available somewhere, or would you be willing to share it, so we can debug it?

Yes so it's not pretty because I was testing a lot of stuff but the most final I have and gives this org.w3id.cwl.cwl.File with cwl-runner/cwltool version 3.1.20210628163208 and schema-salad-tool version 8.1.20210627200047 is here. https://github.com/kbolton-lab/analysis-workflows/blob/master/definitions/pipelines/CH_exome_Final2.cwl

[mr-c]

@Bolton-lab Thanks for the testing document, however when I run cwltool --validate I get an error, so it is making it difficult to confirm if we have fixed your issue or not:

cwltool --validate https://github.com/kbolton-lab/analysis-workflows/raw/403afbbad8d4197166bad8e407e8b99fec0f4f49/definitions/pipelines/CH_exome_Final2.cwl
INFO /home/michael/cwl-upgrader/env310/bin/cwltool 3.1.20211107152837
[many warnings elided]
ERROR Tool definition failed validation:
https://github.com/kbolton-lab/analysis-workflows/raw/403afbbad8d4197166bad8e407e8b99fec0f4f49/definitions/pipelines/CH_exome_Final2.cwl:481:9: Step is missing
                                                                                                                                                required parameter
                                                                                                                                                'reference'

[BJWiley233]

ok ill update it now.

@mr-c it is updated now. just tested and it works with version 3.1.20211020155521

[mr-c]

ok ill update it now.

@mr-c it is updated now. just tested and it works with version 3.1.20211020155521

Great news, thanks!

[jonassmedegaard]

Is this issue perhaps not truly fixed yet?

Using cwl-upgrader 1.2.2 and then cwltool 3.1.20211104071347, both with schema-salad 8.2.20211104054942, on sample file Lancet.cwl, fails like this:

$ schema-salad-tool --print-rdfs Lancet.cwl
/usr/bin/schema-salad-tool Current version: 8.2.20211104054942
Schema `Lancet.cwl` failed link checking:
Lancet.cwl:13:1:  checking field `inputs`
Lancet.cwl:16:3:    checking field `TumorInput`
Lancet.cwl:17:5:      Field `type` references unknown identifier `File`, tried
                     file:///home/jonas/projects/DEBIAN/tmp/Joep/bio-cwl-tools/Lancet/Lancet.cwl#File
Lancet.cwl:23:3:    checking field `NormalInput`
Lancet.cwl:24:5:      Field `type` references unknown identifier `File`, tried
                     file:///home/jonas/projects/DEBIAN/tmp/Joep/bio-cwl-tools/Lancet/Lancet.cwl#File
Lancet.cwl:30:3:    checking field `Reference`
Lancet.cwl:31:5:      Field `type` references unknown identifier `File`, tried
                     file:///home/jonas/projects/DEBIAN/tmp/Joep/bio-cwl-tools/Lancet/Lancet.cwl#File
Lancet.cwl:56:3:    checking field `BedFile`
                     Field `type` references unknown identifier `File`, tried
                     file:///home/jonas/projects/DEBIAN/tmp/Joep/bio-cwl-tools/Lancet/Lancet.cwl#File
Lancet.cwl:255:1: checking field `outputs`
Lancet.cwl:256:3:   checking field `vcf`
Lancet.cwl:258:5:     Field `type` references unknown identifier `stdout`, tried
                      file:///home/jonas/projects/DEBIAN/tmp/Joep/bio-cwl-tools/Lancet/Lancet.cwl#stdout