BIND 10 #1006: loader error for sunstudio build

[BIND 10 Development]

#1006: loader error for sunstudio build
-------------------------------------+-------------------------------------
            Reporter:  jinmei        |                        Owner:  jreed
                Type:  defect        |                       Status:  new
            Priority:  major         |                    Milestone:  Next-
           Component:  build system  |  Sprint-Proposed
           Sensitive:  0             |                     Keywords:
         Sub-Project:  DNS           |              Defect Severity:  High
Estimated Difficulty:  0             |  Feature Depending on Ticket:
         Total Hours:  0             |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------
 I figured out the root cause of the run time failure for the sunstudio
 build:
 http://bind10.isc.org/~tester/builder/BIND10/20110609145000-Solaris10
 -sparc-Sunstudio/logs/build.out

 This is because "-pthreads" was passed to the linker when it builds a
 program that uses liblog4cplus.  '-p xxx' has a different meaning for
 the Solaris (SunStudio?) linker, which is why we saw "threads" in a
 strange ELF section of the resulting binary.

 This flag actually comes from the inherited_linker_flags variable set
 in ~jreed/opt/log4cplus-1.0.4-sunstudio/lib/liblog4cplus.la, and,
 it then turned out to come from the configure script of
 log4cplus.  More precisely, it's because log4cplus/m4/ax_pthread.m4
 sets PTHREAD_CFLAGS to this value unconditionally for Solaris, even if
 -pthreads isn't meaningful for SunStudio (actually it's even harmful
 as we saw).

 A workaround is to rebuild log4cplus with setting PTHREAD_CFLAGS to
 something non null (but that doesn't break the build).

 % ./configure PTHREAD_CFLAGS=" " CXX=/usr/bin/CC

 worked for me.

 We should also note this trap somewhere in our release notes.

 For a longer term we should probably report this to the log4cplus
 developers.

-- 
Ticket URL: <http://bind10.isc.org/ticket/1006>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development