[Patch] Clean up sbumpc, move back _M_buf_size, do not special case unbuffered _M_underflow...

[Benjamin Kosnik]

>Well, actually for sputbackc I applied the scheme you followed for the
>other streambuf inlines: first I split the existing buffered test 1.cc
>in three parts, for in, out and in|out openmodes. In fact the in part
>was missing and I added it(this catched many bugs in my first tries!).
>Then, I created the corresponding 2.x counterparts (2-in.cc, 2-io.cc
>and 2-out.cc), for the unbuffered case, as you did, if I'm not
>mistaken. I think that the sputbackc inline is now sufficiently tested!

Yes, it looks great.

>As far as the pbackfail virtual is concerned (which, by the way, I
>didn't touch at all) I agree that some tests are needed, both for the
>buffered and the unbuffered case. I plan to work on this but _after_
>having prepared the unbuffered counterpart of the seekoff/seekpos
>testcases, if you agree. If you think that some pbackfail tests are
>more urgent, I would be happy to change my priorities, just let me
>know!

No, this sounds like a great plan. I just didn't want you to lose test
cases.

-benjamin