PyUnicode_Append() cannot call directly resize_compact(): I forgot that a
string can be ready *and* not compact (a legacy string can also be ready).

Bug introduced by changesets 7ed9993d53b4 and edf029fc9591.

Bug introduced by changeset 7ed9993d53b4.

 * Add also min_char attribute to _PyUnicodeWriter structure (currently unused)
 * _PyUnicodeWriter_Init() has no more argument (except the writer itself):
   min_length and overallocate must be set explicitly
 * In error handlers, only enable overallocation if the replacement string
   is longer than 1 character
 * CJK decoders don't use overallocation anymore
 * Set min_length, instead of preallocating memory using
   _PyUnicodeWriter_Prepare(), in many decoders
 * _PyUnicode_DecodeUnicodeInternal() checks for integer overflow

registry does not grow exponentially with generation of process.

Issue #17782: Fix undefined behaviour on platforms where ``struct timespec``'s "tv_nsec" member is not a C long.

- Issue #17782: Fix undefined behaviour on platforms where ``struct timespec``'s "tv_nsec" member is not a C long.

registry does not grow exponentially with generation of process.

Thanks to David D Lowe for reporting.

Thanks to David D Lowe for reporting.

  variable if empty path argument is specified.  Patch by Serhiy Storchaka.

  variable if empty path argument is specified.  Patch by Serhiy Storchaka.

The C version tried to remove trailing whitespace between the last quote and
the newline character. I am not sure why it had this because pickle never
generated such pickles---for this to happen repr(some_string) would need to
return trailing whitespace. It was maybe there to make it easier for people
to write pickles in text editors. Anyhow, the Python version doesn't do this
so there is no point keeping this around anymore.

Also, I've changed the exception raised when a bad pickle is encountered.
Again this unlikely to make much difference to anyone though it does make
testing slightly nicer for us.