Re: buildworld (3.4) aborts during compile of libstdc++.so.9

[John Marino <dragonflybsd@xxxxxxxxx>]

On 5/1/2013 13:25, Ralf Folkerts wrote:
> Am 01.05.2013 11:51, schrieb John Marino:
>> On 5/1/2013 11:18, Ralf Folkerts wrote:
>>> Hi,
>>>
>>> I wanted to update my DFly 3.2 System to 3.4. I first moved to the
>>> DragonFly_RELEASE_3_4 Branch, cleaned /usr/obj and issued a make
>>> buildworld. It aborted during compilation of libstdc++.so.9. Then I
>>> cleaned /usr/src and /usr/obj and checked-out the DragonFly_RELEASE_3_4
>>> Branch anew. However, same error (also tried using sh instead of tcsh,
>>> but same error, too). Finally I removed /usr/src and started from
>>> scratch - same error. As bw seems to work for others I must have a made
>>> a big mistake - but didn't find that, yet :(
>>>
>>
>> It didn't build libstdc++ because it couldn't find libm.so.  Assuming
>> that it's linking the just-built libm and not the host version, it
>> would imply that libm isn't getting built either.
>>
>> blowing away /usr/src and reinstalling is fine.  What you did to
>> restore /usr/src looks fine to me.  It means you can be sure there
>> aren't any object files there, a common explanation of problems.
>>
>> I would start by confirming host libm.so is at /usr/lib/libm.so and
>> then I'd check the world build log (e.g. make buildworld |&  tee
>> /tmp/world.build.log) so see what is happening when libm is built.
>
> Hi John,
>
> I already did pipe the build into tee, so I have a log. However, I
> didn't find much libm in, unfortunately.
>
> # grep libm.a buildworld.out
> (snip)

I didn't see the shared version getting built in the log

> In /usr/local/lib, libm.a|so is present
>
> beastie# ls -l /usr/lib/libm.*
> -r--r--r--  1 root  wheel  492234 Jan  4 00:02 /usr/lib/libm.a
> lrwxr-xr-x  1 root  wheel       9 Jan  5 09:56 /usr/lib/libm.so ->  libm.so.3
> -r--r--r--  1 root  wheel  126464 Jan  4 00:02 /usr/lib/libm.so.3
>

What about "find /usr/obj/usr/src -name libm.so\*" ?

Regards,
John