[clang] [clang][analyzer] Improve StdLibraryFunctionsChecker 'readlink' modeling. (PR #71373)

[via cfe-commits]

DonatNagyE wrote:

> Moreover, I think that TC shows that it would be useful to see the behavior of this change on open source projects as I fear that this change could lead to "technically TP, but not helpful for the user" reports.

On a second thought, AFAIK there are already other functions that have similar "argument is zero" corner cases that turn into separate execution paths, so it's unlikely that the impact from this (not too common) function would be especially severe. I'd slightly prefer to see open source results, but if you feel that they're not needed, then don't bother with collecting them.

https://github.com/llvm/llvm-project/pull/71373