On Wed, Mar 1, 2023 at 10:30 AM Thomas Munro <thomas.munro@gmail.com> wrote:
> On Wed, Mar 1, 2023 at 12:09 AM Eugene M. Zheganin <eugene@zhegan.in> wrote:
> > 3) how do I fix it ? Should I take locale sources for ru_RU.utf8 on Linux and compile it on FreeBSD - will it help
?
>
> Out of curiosity (I'm not saying it's a good idea!), do you know if
> FreeBSD's localedef can compile glibc's collation definitions? In
> theory they are in a format standardised by POSIX... I suspect there
> may be extensions and quirks...
Another thought: if you upgrade to 15, you could use ICU as the
default collation provider, and then make sure you have the same major
version of ICU on your Linux and FreeBSD systems (which probably means
compiling from source on FreeBSD, as FreeBSD tends to have a newish
ICU in packages, while popular stable Linux distributions tend to have
a fairly old one). As far as I know, Linux and FreeBSD systems on
amd64 arch should match in every other important detail already (they
both follow the System V ABI, so there shouldn't be any layout
differences in eg structs that are sent over the wire AFAIK).
For what it's worth, for the future we are trying to figure out how to
support multi-version ICU so that you could explicitly set the
provider to ICU v72 to get that sort of thing working across OS
versions and even "close enough" cousins like your case, without
having to compile anything from source, but unfortunately we didn't
manage to get it ready in time for 16.
