Problems cross-compiling Qt/Embedded 3.3 for big-endian architectures:
When cross-compiling on a little-endian machine (e.g. x86) for a big-endian processor (e.g. PowerPC), the configure script will use the host's endianness instead of the target's. Workaround: after running configure, and before running make, edit qt/include/qconfig.h and change the definition of Q_BYTE_ORDER.
Problems converting double to string on ARM Linux:
Some ARM installations have floating point implementations that are not fully IEEE-compliant. In Qt/Embedded 3.3.x, this causes QString::number(double) and related functions to give incorrect results. A fix for this was introduced in Qt 3.3.2: defining the macro QT_QLOCALE_USES_FCVT will make Qt use system library code instead, i.e:
./configure -embedded arm -DQT_QLOCALE_USES_FCVT ...