OpenAlex names the repository copy but its `landing_page_url` is an author-listing page, so the #461 fall-through has nothing to follow

[sotashimozono]

#461 landed in 0.8.10 and the fall-through does ask OpenAlex now. On the shape below it still
returns no OA PDF available, and the reason is not that OpenAlex is silent — it names the
repository, but with a URL that is not an item.

Measured, doiget 0.8.11, all six optional sources enabled

export DOIGET_CONTACT_EMAIL=... DOIGET_ENABLE_CORE=1 DOIGET_ENABLE_DATACITE=1 \
       DOIGET_ENABLE_DOAJ=1 DOIGET_ENABLE_EUROPE_PMC=1 DOIGET_ENABLE_HAL=1 DOIGET_ENABLE_OPENAIRE=1

doiget fetch 10.1109/tsp.2023.3269664
#  (metadata-only: no OA PDF available)  [crossref/closed]

What OpenAlex actually reports for that DOI:

locations: 2
  -    IEEE Transactions on Signal Processing | https://doi.org/10.1109/tsp.2023.3269664
  -    Strathprints: The University of Strathclyde | https://strathprints.strath.ac.uk/view/author/70486.html>

The second location is the institutional repository copy #461 is about. Its
landing_page_url is an author-listing page — /view/author/70486.html — with a stray >
appended, so there is nothing for the fall-through to follow, and is_oa is false on it too.

The copy exists and is one search away:

https://strathprints.strath.ac.uk/85235/
  → .../85235/1/Weiss_etal_TCP_2023_Eigenvalue_decomposition_of_a_parahermitian_matrix.pdf
     (1.66 MB, the accepted manuscript, 15 pp.)

Same for the two siblings — 10.1109/tsp.2021.3049962 → /75128/, 10.1109/tsp.2024.3387726 →
/88703/ — all three with an author-listing URL in OpenAlex and a real item behind a title search.

The control that separates "broken URL" from "no copy"

10.1109/access.2024.3495502, gold OA, same authors, same repository:

  - OA Strathprints … | https://strathprints.strath.ac.uk/91130/7/Khattak-etal-IEEE-Access-2024-Scalable-analyti…

A proper item PDF URL, and doiget fetch gets it. So the field shape differs between the two, and
the fall-through is only usable on one of them.

Two things this suggests, in increasing order of ambition

1. Say what was seen. The found-nothing path (feat(cli): the found-nothing path exits 0 with no trace — report which sources ran, the command that widens the search, and which source is most likely to hold it #505) would have made this diagnosable in one
   run: "OpenAlex named 1 repository location; its URL is not an item page" is a very different
   message from "no OA PDF available", and it points the user at the repository rather than at
   giving up.
2. Follow the repository, not just the URL. When a location's host is a known repository
   platform and its URL is not an item, the platform's own search is a documented, stable interface
   — EPrints answers /cgi/search/advanced?title=…&title_merge=ALL and returns item URLs. That is
   a bigger change and may belong behind a flag; survey(source): trustworthy sources not yet covered — preprint servers, physics indexes, regional platforms #474's "trustworthy sources not yet covered"
   survey is probably where it belongs rather than here.

A trap worth recording either way

Both of my first two search attempts returned zero hits for method reasons, not absence: the
EPrints search endpoint 302-redirects (needs -L) and the DOI field is not id_number. A positive
control on a paper known to be in the repository is what separated the two — without it I would
have reported "not deposited".