On 11/27/2011 02:17 PM, Fraser Adams wrote:
Thanks guys,
however I'm still curious - clearly a release make shouldn't break like this.
I noticed the cast to void e.g. (void)::write(notifyFd, ".", 1); and some
Googling seems to suggest that that should help (though actually using the
return value sensibly seems neater) but I've no idea why the void cast doesn't
work in my Ubuntu system when I assume it does for whoever tested the release.
My C++ is a bit rusty and I'm afraid that I don't know how "declared with
attribute warn_unused_result" is actually enabled in the first place - is it a
compiler switch? I'm guessing it's somewhere in the prototype for "write" that
gets enabled by some preprocessor magic. Perhaps the gcc defaults are different
for Ubuntu than for Redhat?
Clearly it's "work-around-able" but presumably the intention is that
distributions should *just work*.
Most likely you have a different version of gcc which is more sensitive about
the warning. We definitely do want builds to just work but we don't have all the
distro/compiler version combinations represented among the regular qpid
contributers, so things like this can slip out. We need people like you to raise
the issues so we can fix them :) Raise a JIRA and assign it to me.
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:[email protected]
