Re: [oe-core] 3 build machines all failing in the same manner - was:ld segfault cant figure it out

[Phil Blundell <philb@gnu.org>]

On Tue, 2012-04-24 at 13:57 -0700, j wrote:
> The build machines all give messages like this just vary on libc version 
> listed
> 
> [47152.769257] ld[22617]: segfault at 403 ip 00007f2149cfd435 sp 
> 00007fffb83ccb50 error 4 in libc-2.15.so[7f2149cb6000+197000]
> 2012-04-18T14:50:08-07:00 localhost kernel: [47163.860996] ld[31096]: 
> segfault at 403 ip 00007fc2da152435 sp 00007fffab491c30 error 4 in 
> libc-2.15.so[7fc2da10b000+197000]

Does this correlate with an actual problem, or did you just see those
messages in the log and think that they were cause for concern?

If I had to guess I would say that this is probably some autoconf test
which is tickling a bug in ld.  You could probably establish which
package is doing that by looking at the timestamps and correlating them
with the build logs.

> qa.log has more than these messages but every machines logs are the same 
> like this
> 
> udev-182: udev: 
> /home/gusterson/OpenEmbed/setup-scripts/build/tmp-angstrom_v2012_05-eglibc/work/armv7a-angstrom-linux-gnueabi/udev-182-r0/packages-split/libgudev/lib/libgudev-1.0.so.0.1.1 
> links to something under exec_prefix

That doesn't appear to be related to the ld segfault
.
> each one of the machines from a fresh install have this problem. The 
> resulting image boots but does not fully function. I have issues such as 
> the native sdk does not install all the packages correctly.
> example would be install.
> normall you would find it
> /usr/bin/install
> I end up with
> /home/bill/OE/usr/bin/install

... and nor really does this.  This latter problem is almost certainly
just a bug in the SDK packaging and nothing at all to do with your build
host.

p.