I've now merged PR #2982 which fixes all the originally reported ODR violations. If there actually still are any left that I haven't been able to reproduce yet, please feel free to reopen this ticket with very detailed instructions on how to reproduce those errors.

I’m sorry, but I’m unable to reproduce those compilation errors. Would you please be so kind and provide very detailed instructions on how to reproduce them?

When building 75.1 (the layoutex component), I also newly get:

Sure it would be nice. I do understand it's not worth the additional text matrix effort for most projects, so I don't necessarily expect test coverage along with a fix.

Gentoo will almost certainly report a bug after the fact if it ever regresses in a new release, if it helps.

We do have a CI that loops through every distro package and tests this exact issue. That's how we encountered the bug. The cost/benefit for us is different because it is a “tinderbox” to check that the distro is in good working order, with or without random lints (such as GCC + LTO) applied. Nothing is gated on the CI passing, so it doesn't matter if an issue takes several days or a week to discover it, we will eventually find it.

As we are mostly LTO-clean now we're intending to keep on top of new issues.

Clang with -flto -Werror=odr doesn't find these errors as GCC does and there’s no existing ICU4C CI job that runs GCC with LTO, so adding this to the CI would require adding a new job and that’s not necessarily a good idea, for building ICU4C and its tests with GCC LTO takes quite some time and even though it’d be nice we’ll never be able to run every possible test in the CI so maybe this one simply isn’t important enough to add one more job.