As mentioned in D88144, this is only used as a fallback for symbols
missing in our stdsymbol mappings and preventing us from figuring out what's
broken.
After this patch we'll regress some cases, e.g. std::move, but those were
already broken for libc++ or C.
I am also happy with coming up with special casing sfor obvious regressions,
like std::move, before landing. The special casing I have in mind could either
be:
- having a more targeted version of the suffix mapping targeted at those symbols, so we can stop the bleeding as we figure out cases even if we can't invest in a more principled solution.
- just picking a single header in case of ambigious symbols, this might end up being annoying as clangd will insist on inserting the wrong header.
- disable include insertion completely for problematic symbols until we figure out a solution.
should we update these comments?