Would you be willing to also add a small test? Maybe at t/graphics/orcidlink.tex or similar.

And it can depend on the sty file having been installed in t/65_graphics.t:

latexml_tests("t/graphics",
	      requires=>{colors=>'dvipsnam.def',
			 xcolors=>'dvipsnam.def',
                         orcidlink=>'orcidlink.sty'});

Would you be willing to also add a small test? Maybe at t/graphics/orcidlink.tex or similar.

The test would hardcode the specific SVG produced by the current version of orcidlink. Would that cause problems in the future, should orcidlink modify the logo (unlikely!)?

Would you be willing to also add a small test? Maybe at t/graphics/orcidlink.tex or similar.

The test would hardcode the specific SVG produced by the current version of orcidlink. Would that cause problems in the future, should orcidlink modify the logo (unlikely!)?

We may need to update it if/when such transitions happen, which shouldn't be too painful. Probably a small price to pay, so that we also get notified if some future internal change breaks the expected sty load.

Ok, the test is not feasible because TL2021 and later have different implementations. I have updated the binding to make the TL2021 version work, but I have found a major issue with all versions: when \orcidlink is used inside \author{}, LaTeXML omits the SVG colours and you only get a black disc. Not sure why! Converting to draft until somebody can fix it.

The testing harness already has a nice technique to specify a minimal texlive version. Checking this test only for the latest texlive should be fine. That is possible via:

requires=>{ #...
   orcidlink => {packages => 'orcidlink.sty', texlive_min => 2021}}

Figuring out what happens in the \author case is definitely worth a debugging deep dive, especially since one would expect this is where most of the use will come in practice.

In earlier latexml decades I would even be tempted to suggest hardcoding the expected SVG in the Perl binding as a DefConstructor, to guarantee robustness. That used to be common when we had no other easy path forward. But best to check if the raw interpretation can't be fixed first. I'll try to take a look.