TEST-XSPEC.xpl raises error in XML Calabash

[galtm]

The TEST-XSPEC.xpl pipeline works under Morgana, but raises an error under XML Calabash 3 alpha 9.

Fatal err:XS0006: No connection for primary output port ?xspec-html-report?.
   at file:/C:/.../oscal-xproc3/xspec/xspec-execute.xpl:170:23

---

Side question about TEST-XSPEC.xpl and XML Calabash 3: Would having XML Calabash handle the .xspec file name extension similar to .fo as in this table be a reasonable enhancement request for XML Calabash? The first bullet of https://github.com/usnistgov/oscal-xproc3/blob/main/tutorial/sequence/Lesson02/walkthrough_102.md#test-xspec explains the workaround you had to do, and it seems that the default mappings in XML Calabash are trying to smooth that out. (Maybe you'd still need your workaround for Morgana, though.)

[xml-project]

There is a bug in MorganaXProc's static analysis: A missing connection for a (primary) output port is not correctly identified during static analysis. There should be an XS006, as correctly raised by XML Calabash 3.
MorganaXProc-III currently raises dynamic error XD0007 if the respective subpipeline is executed. This might not always be the case, e.g. in a p:choose where the defective alternative is not executed.
Hope that helps,
Achim

[wendellpiez]

@xml-project yes it helps! thanks very much. In a rewrite we'll try to find a solution that works in both processors, or a stopgap which we can roll back when the Morgana bug fix is in. 🥬

[xml-project]

@wendellpiez Any correct pipeline (accepted by XML Calabash) will also work with MorganaXProc-III (present and future).

[wendellpiez]

@xml-project thanks, that is helpful to know.

[galtm]

Thanks, @xml-project and @wendellpiez !

[wendellpiez]

Where the XProc has

oscal-xproc3/xspec/xspec-execute.xpl

Line 171 in bb76ea1

<p:output port="xspec-html-report" primary="true"/>

making it

<p:output port="xspec-html-report" primary="true">
  <p:empty/>
</p:output>

Appears to satisfy XML Calabash.

I.e. it is just a bit more scrupulous than Morgana. (Good lesson.)

An alternative design could use p:error instead of failing silently, as only a non-functional XSpec will hit the p:otherwise (tmk).

[galtm]

An alternative design could use p:error instead of failing silently, as only a non-functional XSpec will hit the p:otherwise (tmk).

You're right, only a non-functional XSpec would hit that. The XSpec compiler would raise an error if <x:description> did not have any of the attributes stylesheet, query, or schematron (link to 3 XSpec compiler files, one per language).

[wendellpiez]

Let's close the Issue as Done For Now (we can revisit later if we like).