Patch for PR 10728

[Tom Tromey]

Chris> Is there a cleanup for stuff that gets done in
Chris> _initialize_cp_valprint()?

There may be, but I wouldn't bother with it.
It is ok to initialize an obstack and then leave it around forever.

Chris> The problem is I don't know which eps are relevant.  Does anything
Chris> other than C/C++ have the statics problem?

Yeah, Java does at least.  I'm not sure about other languages, but I
assume so.

However, Java doesn't call into the code you're modifying, so it
shouldn't matter for this patch.

>> The formatting is weird here; the first statement doesn't appear to have
>> changed, but has gratuitous formatting changes,

Chris> I just thought it made the stmts more readable.

Usually we recommend formatting changes as separate patches.  Personally
I'm more inclined to just leave bad formatting alone, to avoid adding
noise to the history, but that's just me.  Also, the result still has to
conform to the GNU standards.

Tom