Closed larsoner closed 3 weeks ago
And the bug is really in Sphinx.
Is there a reference for this somewhere? It triggered under pretty innocent-looking circumstances, so it'd be a pity IMO to have no mechanism in place to diagnose, or at least a sphinx issue to refer to.
Yep, just took me a while to write up!
Thanks for digging into this issue across scipy/PST/sphinx! 🙏
See https://github.com/scipy/scipy/pull/20897. TL;DR is that the problem was always with reading in parallel, not writing. And the bug is really in Sphinx.
It would be good to detect and clean this up in PST somehow perhaps as well -- maybe with a warning if a top-level / root page ever shows up as a child? This might only be an issue when building in parallel, still not totally clear to me why things were okay in the serial case.
I think any additional changes could be made in a follow-up PR probably.
closes #1643