This fixes the build problem that I described in #212
(NOTE: There's actually one other un-#if-guarded usage of the variable "fUseUPnP" in this file, but that one doesn't seem to cause build problems, for whatever reason. So I just added #if guards around this usage, to make the change as minimal as possible.)
This is not correct. The point of fUseUPnP is that it should always evaluate to false if upnp is absent. That puts the #ifdef in a single location, rather than requiring an ifdef at every site, crapping up the code.
ah -- perhaps ths change should actually be "#ifdef", not "#if"?
It looks like fUseUPnP is only defined if USE_UPNP is defined -- quoting main.cpp:
<pre> #ifdef USE_UPNP #if USE_UPNP int fUseUPnP = true; #else int fUseUPnP = false; #endif #endif </pre>
...though it's declared unconditionally in main.h:
<pre> extern int fUseUPnP; </pre>
Maybe "fUseUPnP" should be defined & set to false in main.cpp, if USE_UPNP isn't defined?
Probably should be, though it doesn't cause issues unless you have stale outputs in obj/*
OK, this latest commit reverts my original changes, and instead groups the "USE_UPNP undefined" situation in with "USE_UPNP=0" situation, for defining fUseUPnP.
Probably should be, though it doesn't cause issues unless you have stale outputs in obj/*
It actually did cause issues for me, the first time I ever built bitcoin.
What compiler version? The only ones I've ever seen don't cause problems there. Did you not start to compile with USE_UPNP=0/1 then stop and go back and compile with USE_UPNP=
Did you not start to compile with USE_UPNP=0/1 then stop and go back and compile with USE_UPNP=
Sorry, yeah -- you're totally right. I thought I was building from scratch, but I think what happened was:
So this indeed only seems to be an issue when you build with stale object files in /obj. Still, I think the above series of steps are probably not uncommon, among people who lack the upnp library, so I think it's a build use-case that's worth being concerned about. :)
Well, it doesn't matter that much as the autotools stuff should fix all this.
Can someone roll these into a single commit, please? It is ready to pull, otherwise.
We prefer not to pull in situations where we're pulling "commit A, now commit B fixes commit A" type situations, to avoid cluttering the upstream bitcoin git history.
I can do that, sure - should I just do that as a fresh commit & fresh pull request, then? (sorry, I'm new to github)
Did it for you :), fUseUPnPfix branch, ACK.
ah, thanks! :)
pulled bluematt's fix