[ ] I don't know how to test a pandoc3 figure - where does this arise? I fixed this but didn't test it.

If you have an image with a caption but without a crossreferenceable ID, it should be a pandoc3 Figure. Something like

---
format: typst
---

![Hello]({{< placeholder >}})

• Similarly, what is a PanelLayout in Typst? I tried the HTML Panel Layouts ::: {layout = "[ [1,1] ]"} but no surprise that didn't work.

I believe you that it didn't work, but we do have support for it, in

Thanks @cscheid! I was able to test pandoc3 figure with

![Hello]({{< placeholder >}})

and the subfloat case case for PanelLayout with

::: {#fig-panel layout-ncol=2}

![Marcador]({{< placeholder 200 >}}){#fig-layout-a}

![Marcador]({{< placeholder 200 >}}){#fig-layout-b}

Panels

:::

IIUC the final case of make_typst_figure() in layout/typst.lua should be a PanelLayout without crossreferenceable content:

::: {#fig-panel2 layout-ncol=2}

![Marcador]({{< placeholder 200 >}})

![Marcador]({{< placeholder 200 >}})

Panels

:::

But anything I put there, I get the inscrutable error

Block, list of Blocks, or compatible element expected, got Blocks
	while retrieving function argument content
	while retrieving arguments for function Div

stack traceback:
	...c/quarto-cli/src/resources/filters/./ast/customnodes.lua:527: in local 'scaffold'
	...on/src/quarto-cli/src/resources/filters/./ast/render.lua:46: in local 'filter_fn'
	...c/quarto-cli/src/resources/filters/./ast/customnodes.lua:154: in function <...c/quarto-cli/src/resources/filters/./ast/customnodes.lua:144>
	(...tail calls...)
	[C]: in ?
	[C]: in method 'walk'
...

(I thought that Blocks was a list of Blocks!)

I tried to debug this, but the line numbers and functions in the stack trace don't quite make sense to me. Perhaps it merits a separate issue.

It renders as I would expect in HTML

and I think this confirms that the supplement is unused and does not need to be localized.

(I thought that Blocks was a list of Blocks!)

This is a really bad error message from Pandoc, and it can come up in situations where a Blocks list has a non-Block element inside it. It is emitted by Pandoc in their Lua-to-Haskell marshalling code, and it's hard for us to guard against it in general :(

Very helpful clue. Okay, I'm opening a new issue for that one as it doesn't seem like this example has worked right since 1.5.0 but it started crashing with subfloats.

I think this is complete except for looking at the future impact of the fix and submitting a PR on Pandoc. That seems to require a different change to the template but I am not sure that I understand their i18n system yet.

I've pushed the fix to #9854 to this branch, since the test here triggers that.

Ok to merge? When we get a fix upstream in Pandoc we can revert to their template.

@gordonwoodhull I don't think we should have that subfigure situation - if we're going to have supplements and numbering in that case, then they'd need to be subfloats.

Completely agree, but I think it's outside the scope of this PR. I think we have a small constellation of issues around figures without ids in Typst.

Could we merge this and I'll open another issue with the numbering problem?

Continued in #9857

Squashed and merged to main in d40b8ec, closing this.