currently gcc 4.8.1 on Arch (07.10.13)
Flags in makepkg.conf
CFLAGS="-O3 -pipe -march=armv7-a -mcpu=cortex-a9 -mfloat-abi=hard -mfpu=neon -fstack-protector --param=ssp-buffer-size=4"
CXXFLAGS="-O3 -pipe -march=armv7-a -mcpu=cortex-a9 -mfloat-abi=hard -mfpu=neon -fstack-protector --param=ssp-buffer-size=4"
PKGBUILD needs one patch adding (neon-update3.patch) and one removing (armhf-triplet-trunk.diff)
As it is patch will fail to find the file to be patched and you will have to supply the full path.
in my case
Also need to change flags pass to GCC by configure.
[[ $CARCH == "armv7h" ]] && CONFIGFLAG="--host=armv7l-unknown-linux-gnueabihf --build=armv7l-unknown-linux-gnueabihf --with-arch=armv7-a --with-float=hard --with-fpu=neon"
The PKGBUILD with the changes is attached as is neon-update3.patch
The original PKGBUILD is from https://github.com/archlinuxarm/PKGBUILDs/blob/master/core/gcc/PKGBUILD
neon-update3.patch is from http://gcc.1065356.n5.nabble.com/PATCH-ARM-Fix-unrecognizable-vector-comparisons-td947064.html
Sound under linux and KDE seems to able to get itself confused with its settings easily enough so their is no point adding to the confusion. I’m sure that HDMI is hugely useful to those that want it and use it but up to now I never have.
Among the issues I’ve noticed is that at seemingly random times the KDE mixer Kmix will revert to setting the Nvidia HDMI output as the master channel. Then I have to change it back which starts to get old fairly rapidly. Continue reading
I use keys to connect so its not an ssh-askpass issue, though I’ve seen it mentioned elsewhere that some other people have the same issue with connecting using passwords and may have misdiagnosed why the fix works.
virt-manager run as root connects just fine. Continue reading
When using the the Nvidia blob the Chromium (and Chrome) browser leaves zombie processes behind causing the system to slow over time. This is apparently caused by nvidia threads forking when according to POSIX specs they probably shouldn’t, and Chromium’s custom malloc (tmalloc) not playing nicely with this out of spec behaviour whereas the glib malloc does cope with it !