It looks like this issue was already fixed upstream, or at least
partially. I reported a separate bug that also references the upstream
patch and gives some more context:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1057630
While this is a relatively severe bug (uninitialized memory), it doesn't
manifest itself on many architectures.
The build failure that occurs on arm64 and many others is actually a
mistake in my Catch2 v3 patch: I used catch2::WithinRel comparison
everywhere, but this doesn't work well when comparing to zero. The only
surprise here is that it *doesn't* fail on amd64.
I will reopen https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1054697
and submit an updated patch that uses catch2::WithinAbs for comparison
with zero instead.