AaronBallman wrote:

> > I don't think we should play whack-a-mole trying to find all the places we 
> > could potentially print a large integer value and update the diagnostic 
> > call sites. I think this is something the diagnostics engine should handle 
> > when deciding how to render an integer so that the behavior is consistent 
> > across diagnostics. Otherwise we'll have to do the same dance for 
> > `note_constexpr_new_too_large`, `err_attribute_size_too_large`, and so on.
> 
> Can we just enable that by default for all diagnostics? Or is there one where 
> that’s obviously wrong for some reason?

It's not that it's wrong, it's more that it can be distracting depending on the 
user and the situation. e.g. `1'234'567` is easier to read than `1234567` but 
not everyone will agree that `1'234` is easier to read than `1234`. There's 
also the problem that we single quote syntactic constructs and so if you end up 
with some cases like `'1'234'` the separator can be harder to read.

That said, I'm not certain what a good heuristic is, but "always use 
separators" isn't it. Also, we emit hex values sometimes, that might have a 
more obvious heuristic in some ways than decimal values.

https://github.com/llvm/llvm-project/pull/211785
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to