For quite some time, running AutoDoc() without the "autodoc" option was broken. We should have a test case to detect that. Likewise we should have test cases for other combinations of options being present / missing, to catch if this breaks again in the future.
In addition, it is bad if our changes break how existing packages are using AutoDoc. E.g. I am using AutoDoc() in the nq and polenta packages, but without the "autodoc" option, hence it is impossible to build their manual with various "bad" AutoDoc versions.
Thus, I think it would make sense if we collected together all packages that are using AutoDoc in one way or another, and created a batch of tests for these and their documentation -- that is, we keep copies of their docs, and the tests rebuild them, and report a failure if there is any change in the output -- be it an error while running AutoDoc(), or be it a change in the generated HTML documentation.
For quite some time, running AutoDoc() without the "autodoc" option was broken. We should have a test case to detect that. Likewise we should have test cases for other combinations of options being present / missing, to catch if this breaks again in the future.
In addition, it is bad if our changes break how existing packages are using AutoDoc. E.g. I am using AutoDoc() in the nq and polenta packages, but without the "autodoc" option, hence it is impossible to build their manual with various "bad" AutoDoc versions.
Thus, I think it would make sense if we collected together all packages that are using AutoDoc in one way or another, and created a batch of tests for these and their documentation -- that is, we keep copies of their docs, and the tests rebuild them, and report a failure if there is any change in the output -- be it an error while running AutoDoc(), or be it a change in the generated HTML documentation.