svn:mergeinfo props in the trunk?

[Gary Funck <gary at intrepid dot com>]

After merging a recent version of the trunk into the gupc branch,
I noticed that the branch seemed to pick up an svn:mergeinfo
property from the trunk.  My general understanding is that
checking svn:mergeinfo property values back into the trunk is
discouraged.  For example:

On 06/24/10 11:28:17, Joseph S. Myers wrote:
> [...] I think some people have expressed a dislike of svn:mergeinfo
> properties referencing random branches being added to trunk. [...]

Yet, there appear to several stray occurrences:
$ svn up
At revision 167969.
$ svn pg -R svn:mergeinfo
gcc/testsuite/go.go-torture - /trunk/gcc/testsuite/go.go-torture:154083-167382
gcc/go - /trunk/gcc/go:154083-167382
gcc/testsuite/lib/go-dg.exp - /trunk/gcc/testsuite/lib/go-dg.exp:154083-167382
gcc/testsuite/go.dg - /trunk/gcc/testsuite/go.dg:154083-167382
gcc/testsuite/go.test - /trunk/gcc/testsuite/go.test:154083-167382
libjava/classpath - /branches/CLASSPATH/libjava/classpath:164526
gcc/testsuite/lib/go.exp - /trunk/gcc/testsuite/lib/go.exp:154083-167382
libgo - /trunk/libgo:154083-167382
gcc/testsuite/gcc.target/powerpc/ppc-round.c -
/trunk/gcc/testsuite/gcc.target/powerpc/ppc-round.c:165844-166448
gcc/testsuite/lib/go-torture.exp -
/trunk/gcc/testsuite/lib/go-torture.exp:154083-167382

(some line-wrapping occurred above)

I recall that Ian had indicated that he wanted to keep the 'gccgo'
branch open for potential experimentation/test before merging
back into the main line.  That might explain the occurrences
of those svn:mergeinfo properties?

Is there a policy regarding svn:mergeinfo properties in the trunk?

If it is OK for these properties to appear in the trunk,
should they be preserved when merging the trunk into a branch? 

thanks,
- Gary