Closed gdt closed 3 years ago
did you have an error during the build? unicode_util_compat is used as an abstraction so it should be there.
Next version won’t include it. See 230a9175a5edb1b1a47fb09af2f4341eab93b1b0 but in hte version yoûre using it should be there which version of unicode_compat did you install?
I didn't install any version of unicode_utils_compat because the package (in pkgsrc) doesn't depend on it, which is I think because the README.md doesn't say anything about it being needed.
However, the compat build objects didn't get installed because the build recipe wasn't expecting them. With that adjusted, the examples work. I am unclear on whether there is the equivalent of "make install" for erlang/rebar; from looking at how pkgsrc handles erlang modules it seems there isn't. But that's not an erlang-idna issue.
It's good unicode-utils_compat won't be included. In pkgsrc, we don't allow builds to use the network. A new release would be much appreciated.
With the package build adjusted to include the unicode compat build products, ejabberd s2s and startup is working again. Thanks for explaining.
I ended up here after a package upgrade, wtih ejabberd complaining that plain-ASCII domains (for s2s) were not valid IDN. Beware that I know very little about erlang.
I'm using erlang 23.3.4 (V11.2.2 from erl +V) on NetBSD 9. I have been successfully running ejabberd, so things are at least 99% ok.
With erlang-idna 6.1.1 (the latest release, as I read the tags), I get exceptions running the example from the README:
And indeed I do not have that module. The README for idna doesn't mention it, and it seems that it is needed for older erlang. Given that unicode_util seems to be normal in up-to-date erlang, I don't follow why that isn't used, with the compat used as compat if necessary.