On Wednesday 02 January 2008, Bernhard R. Link wrote: > * Sune Vuorela <firstname.lastname@example.org> [071231 17:26]: > > If people want to create patches for the unaligned load/store they are > > most welcome to do so. I don't plan to. > > Please note that there are architectures where emulating unaligned > accesses is not possible (AFAIK at least sparc), many architectures > also have this with some special accesses (I heared rummors that > MMX access on i386 also needs aligned access, so a future better > optimizing gcc may also cause it to fail on i386). it hasn't given problems on any archs except hppa. and linux-hppa isn't one of the quite many archs that is supproted by upstream. It should be quite easy to locate the parts of the code that does the unaligned access in qt4. Last I looked, it ended up in qt4-x11-4.3.3/src/corelib/global/qnumeric_p.h with lines like this: (61,74,87) return *reinterpret_cast<const double *>(QSysInfo::ByteOrder == QSysInfo::BigEndian ? qt_be_inf_bytes : qt_le_inf_bytes); > For all of those reason I personally consider any buildd setup broken, > where such catchable unaligned does not cause a bus error. (That is only > my personal opinion, though I personaly tend to get very personal when > discussing this). Then please don't spend too much time debating it, but just do a patch for qt4 instead ;) /Sune -- Genius, I cannot send the RW display, how does it work? The point is that you can never log from a front-end of a prompt to the digital proxy to a PCI display on the DirectX monitor over a RO utility.
Description: This is a digitally signed message part.