pthread7-rope fails also for std::string :(

[Benjamin Kosnik]

>Ah, ok! Honestly, I haven't studied in detail that report: actually is a 
>*very* serious
>issue, a blocker for 4.0, in my opinion! Perhaps you should clarify this 
>in the PR!

We are of the same mind, my friend.... I set it to critical and Wolfgang
confirmed.

>This is definitely not needed: to the contrary, "poor" rope ;) will 
>remind us that this
>very serious C++ issue *must* be fixed before 4.0!! Maybe, we could also 
>add to the
>testsuite specific testcases using something else besides rope.

FYI for the rope issue, just set rope_type to use
__gnu_cxx::new_allocator<char>. That sidesteps the other issues.

Then, just run the binary under valgrind.

I think it may be a testcase issue, but I dunno, and didn't have enough
time to look at this closely.

-benjamin