Fix references resolving

[mvorisek]

fix part of #56 - fixes absolute references (but declarations still cannot currently contain \)

also fix #58

also missing targets are newly reported, as infos only, as some targets might be in other projects, global PHP classes, ... and CI must pass

also the testing is unified and the html output for all original tests is asserted using CI

also the code is reformatted, in separate commit, using Black to enforce strict coding style

output of all rst test files is exactly the same as before this PR -> no BC break

[mvorisek]

@markstory sorry for a lot of noise :), but I had to fix a lot of small issues in order to pass all the added tests. This PR is finally done, more to come in other PRs.

[markstory]

This is a big change, it might take me a few days to get through it.

[markstory]

Is this to simplify build configuration? Having this markdown/rst hybrid format be the default isn't ideal to me. It makes a less-frequently used package a hard dependency and requirement for long term maintenance.

[mvorisek]

Sphinx requires only one file format. To allow to test everything at once, we need md files + {eval-rst}.

• MyST parser needs to be tested to assert the parsing is done solely by Sphinx and the phpdomain does not operate on raw rst. This is related with Native directive params support #55 (I plan to fix it in another PR).

• {eval-rst} seems to work perfectly as native rst, Sphinx simply does not read rst file directly, but instead the wrapped string. I did not experience any difference between native rst file and rst from evaluated string.

MyST parser is very popular and the dev/test dependency should not be a problem.

[markstory]

MyST parser is very popular and the dev/test dependency should not be a problem.

Fair enough, this is the first time I've encountered it.

[markstory]

What's up here?

[mvorisek]

:returntype: SOME_CONSTANT return type reference with constant is unresolved (has no clickable link) - I need to find a solution first. If ok, let's merge this big PR and these small TODOs can be fixed later, probably together with #56.

[mvorisek]

@markstory the changes are not smaller, but the most of the changed LoC are changes from generated expected html and changes due Black CS. If you want the the Black CS changes as separate PR, let me know, but as I see you merge PRs using merge, thus the individual commits will be preserved even if merged as one PR.

[markstory]

I had to turn off tests for python 3.9+ as they started failing. I've also turned off the cron schedule for CI as it has been failing for a few days, likely due to a change in sphinx.