On Saturday 23 February 2008 21:45, Paul Floyd wrote: > Looks like I've got there.
Yay! > 1. edit qmakespecs/solaris-cc/qmake.conf to add -library=stlport4 to the > CC and LINK variables. As Stefan points out, we (KDE-Solaris) use rwstdcxx, so there may be real issues there with the templates (since the error I was referring to was probably std::pair related). To get the same compile environment as I am using, though, would require that you get KBE and the FOSS packages built (from CVSDude). I'm sorry you had to go through this rediscovery period: I blogged about it here last week: http://people.fruitsalad.org/adridg/bobulate/index.php?/archives/524-Prepwork-for-FOSDEM-port-Qt.html > 2. Error in Makefile for qatomic.o. Edited the makefile and deleted the > 2nd rule. Deleted qatomic.o, rebuilt. Yup. Thiago doesn't know why this happens either. > 3.Numerous (20 odd) bugfixes in the code, predominantly in the 3rdparty > GCC/Linux cruft. Mostly operator+ string ambiguities. Link errors with > const/non-const overloads. A couple a std namespace problems. Lots of little things like that, yes. > 4. Quite a few link errors with demos/browser, due to functions having > const function arguments either in declarations or in definitions, but > not both. This is common simply because gcc does not complain about it; it's hard to say which fix is right though -- adding const to the header (changes binary compatibility (if it had compiled in the first place)) or removing it fromthe implementation (might lead to bugs later if people are expecting the compiler to catch accidental non-const behaviour). Anyway, if you have stdcxx, can you give it another shot? [ade]
