The documentation for
Tcl_SetObjLength(objPtr, newLength)
suggests that it will leave
objPtr->bytes pointing to a buffer
of at least length newLength+1,
that it will write a null byte
at objPtr->bytes[newLength], and
that it will leave objPtr->length
with the new value of newLength.
This is often true.
However, when objPtr points to
a pure unicode object with at least
one multibyte character, or with
some other history that leaves
objPtr->bytes == NULL and
stringPtr->hasUnicode != 0, then
Tcl_SetObjLength() manipulates
the unicode representation, growing
it to hold at least newLength *characters*
(not bytes). It's possible that a
caller expecting the documentation to be
true might get a nasty surprise.
The ability to grow the unicode rep
is useful. Perhaps that functionality
should have its own API, rather than
having undocumented, semi-magical modes
of operation?
|