Hi Collin, > This patch fixes the two examples that you gave. I think it may be > missing the case where the buffer doesn't end in a partial line > though.
There are probably some other cases that are not handled either. The code is ca. 200 lines of considerations of columns and widths here and there. > I have a feeling it is necessary, but I don't have a test case to make > sure I don't break things. So, my feeling would be that the first thing to do is to create a new unit test for it: 1. so that we don't need to build and try GNU tar in order to see whether there is a regression in the future, 2. so as to cover the tricky cases as well (such as, near the end of a line, a word that ends in 'Ä' or in some width 2 character). Bruno