Hi, To stress-test GCC and LM32, I tried to run Qt4 applications a bit. Attached is a work in progress and kludgy patch for Qt. + configure Qt with ./configure -xplatform unsupported/rtems-g++ -no-qt3support -embedded lm32 -big-endian -host-little-endian -no-largefile -static -no-javascript-jit -no-webkit -no-script -debug -no-separate-debug-info + some mmap calls were savagely removed and the corresponding functions won't work; grep the patch for #warning to find them. + no framebuffer or mouse driver + only the examples/network/download application has the required #defines and init code to get RTEMS working + the linking of the applications does not work yet. Re-run the linker manually by removing -ldl -lpthread and adding -B$RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems + there might be some remaining junk in the patch, I cleaned up most of it but it's pretty hard to remove all the files Qt creates (make *clean only deletes part of them) + with GCC 4.5.1 (binary rtems4.11 toolchain from the RTEMS website) the application crashes at startup because this line of code (qcoreapplication.cpp:552): d->createEventDispatcher(); actually calls the class's destructor instead of the method because of some compiler or runtime fuckup. You can expose this error in GDB by putting a breakpoint on this line, then using print d->createEventDispatcher which will print the correct address of the method to call. Then, use the "disassemble" command to locate the call instruction, put another breakpoint on it, continue execution until the call instruction and use "info registers" to print the jump address... which isn't that of createEventDispatcher(). I do not know the root cause of the problem, but I guess there may be other issues like that in the binary.
Bottom line: C++ support in LM32 looks messed-up. S.
qt-rtems.diff.bz2
Description: application/bzip
_______________________________________________ http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org IRC: #milkym...@freenode Twitter: www.twitter.com/milkymistvj
