On Wed, Apr 23, 2014 at 10:02 PM, Wladimir <laanwj@gmail.com> wrote:
On Thu, Apr 24, 2014 at 12:28 AM, Gregory Maxwell <gmaxwell@gmail.com> wrote:
> On Wed, Apr 23, 2014 at 1:39 PM, Warren Togami Jr. <wtogami@gmail.com> wrote:
>> If you are
>> a rare user who needs Bitcoin-Qt on an incompatible system you can at least
>> build it from source.
>
> Tails users usually can't really build it from source— talks is a live
> boot mostly stateless linux distribution for privacy applications.
> It's really good in general.

Aside: But is Bitcoin Core a well-suited application for those uses? I
cannot imagine someone running a full node on a stateless system.

Anyhow: As this is only one symbol, we can probably get rid of it (as
we didn't use it in 0.8.6?), or put it behind some #ifdef
COMPATIBILITY_BUILD...

Another option: Instead of statically building it'd be easy enough to
build against the 4.6 Qt headers instead without even swapping the
library. Qt is, after all, forward-compatible - between the 4.x
versions. This will lose some GUI features but if compatibility is
more important here that's a choice that can be made.

Wladimir

I now see how it worked with Bitcoin 0.8.6.  Lucid has qt-4.6.2.

It is more than one symbol.  It does not seem to be a wise thing to replace functions beyond the trivial in glibc and libstdc++.

I personally think we need to decide upon a cut-off point beyond which it makes no sense to add the risk of increased complexity.  RHEL6 had qt-4.6.2 as well and I don't think I've heard a single complaint about bitcoin-qt being broken there given almost nobody uses it as a desktop.

Warren