Since no one seemed to be interested in the NUL-separator bugs I've taken
a look at the code myself and I've managed to locate the bugs. 

The function tiedarrsetfn() does not convert characters for which imeta()
is true into separator strings properly. This means that the bug is not
limited to NUL-characters but affects all characters for which imeta() is
true (e.g. 0x83). 

There are also 2 bugs in zjoin():
1. zjoin() doesn't put the NUL-terminator in the right place when delim is
a meta-character.

2. zjoin() refuses to add 0-delimiters (although it does count them when
computing the length). This behaviour does not seem to be used anywhere in
zsh code AFAICS and even if it were, that would be a bug in the respective
call.


The attached patch fixes both problems and adds 2 regression tests to
B02typeset.ztst (both tests are necessary, since 0 and 0x83 are treated
differently in some places, so that one case could regress independent of
the other).

MSB