Feature Offer: LLVM and Clang

[colematt]

Summary

• Dash needs Clang and LLVM docsets targeted towards Clang and LLVM developers, not the compiler front-end users.
• I have figured out how to generate these docsets on each LLVM tagged release, but the .tar.gz of the docset resides on a CD runner and can't be hosted as a Dash feed. I need help hosting these docsets.
• Is there still present demand for this effort? It is non-trivial to generate the docsets because of the size of the docset and the amount of indexing that has to occur.
  • This appears to be a long-standing need, see update LLVM 6.0.0 #1786, update Clang 6.0.0 #1787, Clang and LLVM outdated #2715, Clang 11 Docset #2722, LLVM 10 Docset #2723, etc. But these requests are now 3+ years old and I haven't seen requests since.

---

LLVM's development cycle is a major release approximately every six months, and the online docs represent the most recent release candidate. Search engine hits inevitably point to this documentation, which is actually too new for the typical developer who has probably acquired a libClang/libLLVM major release from their package manager. Furthermore, whoever previously set up the LLVM and Clang documentation actually pointed it at the end-user HTML pages and is scraping them from LLVM's docs site as opposed to generating the API's Doxygen docset after each major release.

Unfortunately, the LLVM docset is around 13 GiB uncompressed and needs about a day of continuous run time to generate on a modest Mac desktop. I would have to set up a self-hosted runner to do it continually because both Github and Gitlab do not provide hosted runners with enough resources. Furthermore, until I find a Linux equivalent to docsetutil, it has to run on a Mac -- I'd love to put this on my university's HPC cluster and get it done an hour or two, but there's no Linux tooling that I know of.

Also, per this repo's README, we will have to make arrangements to get the tar.gz to a maintainer separate from a pull request, it will certainly exceed Github's 100MB limit.

[Kapeli]

Unfortunately I can't handle docsets that large currently. Sorry!