Boost's broken integration with ICU is one problem. Boost detection of ICU generally fails because of an incorrect path assumption, and various versions of boost have smallish bugs that we need to avoid. Sometimes getting a specific version of boost, with static and/or dynamic builds, and multithreading support, and proper ICU integration can be hard. It's easier just to build the correct version with the intended configuration.

I wouldn't qualify the breaks as bad in the sense that they are extensive, just that they need to be avoided.

e

> On Jul 6, 2016, at 7:41 AM, mlmikael <mlmikael@???> wrote:
>
> Eric,
>
> So previously you clarified that the ICU library bundled with various OS:es have been badly broken, and that Boost's consistency check not really would detect the issues, and that the ICU issues out there have been too severe and subtle for LibBitcoin to be able to implement a proper consistency check also - and for this reason you simply make a local compile of ICU for LibBitcoin so you know the ICU version delivers.
>
> And, users can use any other ICU install they want but under the condition that they know it's a really good one.
>
>
> Now - you also make a local compile of LibBoost for LibBitcoin.
>
> What have been the experiences and reasons that led you to do that?
>
>
> Thanks!
> Mlmikael
>
> _______________________________________________
> Libbitcoin mailing list
> Libbitcoin@???
> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/libbitcoin