back
2 comments
Is it possible to avoid thus issue by using the C(root) .UTF8 locale? I am hoping it sorts according to unicode codepoint number, and hence is stable if you unicode-normalize your data.
That's right - the C locale would be immune to this. We are primarily seeing this while migrating customers into Crunchy Bridge whose databases are using language/region specific collations that pre-date the glibc 2.28 updates and either are still running old glibc versions or have upgraded in the intervening years without detecting/addressing the mismatch.