Is exact count of other implementations useful in this error message?

[scottmcm]

I got the following CI failure in #102862

---- [ui] src/test/ui/try-trait/bad-interconversion.rs stdout ----
diff of stderr:

16	             <f64 as From<i16>>
17	             <f64 as From<i32>>
18	             <f64 as From<i8>>
-	           and 67 others
+	           and 68 others
20	   = note: required for `Result<u64, u8>` to implement `FromResidual<Result<Infallible, i32>>`
21	
22	error[E0277]: the `?` operator can only be used on `Result`s, not `Option`s, in a function that returns `Result`

I don't know if this is a diagnostics request or a UI test normalization request, but getting UI test failures when adding a From implementation is kinda annoying (if admittedly easy enough to fix).

So I wonder how much value the exact number here is providing. I wonder if it might make sense to change the message to

and many others

if it's above some threshold.

I can't see any reason the exact count is useful to the programmer reading the error (at least when it's a large number), and it might save some work in error message generation to stop once it's found a bunch.

[compiler-errors]

I'm fine with either of the suggestions, choosing some other phrasing for large numbers of trait candidates or normalizing it in the UI tests themselves.

Coincidentally, @mejrs just implemented a trybuild normalization for this in dtolnay/trybuild@ea3f207.

The code we use to generate "... and {} others" is kinda scattered all over the place for different suggestions, but if we're gonna change one we should change them all.

and it might save some work in error message generation to stop once it's found a bunch.

But this shouldn't be a primary motivation, since we calculate the list of candidates (and other lists used in suggestions that end with "... and {} others") eagerly, and it'd probably be more work to make it lazy -- especially because diagnostics code isn't focused on speed.