[uClibc] Bad PPC toolchain from BuildRoot
ralphs at netwinder.org
Wed Jun 15 17:12:55 UTC 2005
Marc Leeman wrote:
> It seems that we're on the same mailing lists but to get the record
> 1. gcc 3.4.3, created with buildroot compiles u-boot, it does not boot
> the system. I haven't had the time to look into this for our
> boards. We use Wolfgang's toolchain for compiling Das U-Boot.
I'm using buildroot gcc 3.4.3 to compile u-boot for PPC here and it both
builds and runs. Admittedly it is an older version of uboot (1.0.2) and
it has seen all sorts of hacking... but those hacks went in years ago,
while I only switched to gcc 3.4.3 a month ago.
> 2. gcc 3.4.3 compiles the kernel and the kernel boots and does what it
> is expected to do without any hickup. I've been testing this for
> over 5 months on both 2.6.x kernels and 2.4.x kernels (all released
> kernels since 2005).
I'm using same toolchain to build 2.6.11 and userspace, only have found
one issue with the toolchain. I have one odd case where removing a
printk() causes breakage in a driver, but it works fine with it in.
I've looked at generated code, it differs but my PPC asm is not strong
enough to conclude if its a gcc bug or not.
> 3. in our experience, gcc 3.4 creates faster PPC code when compared to
> the 3.3 compilers.
Can't say, haven't run any benchmarks on this.
More information about the uClibc