view nt/TODO @ 30833:2db6e42a6ba3

(MINIMUM_CONVERSION_BUFFER_SIZE): Macro deleted. (conversion_buffer, conversion_buffer_size): Variables deleted. (get_conversion_buffer): Function deleted. (struct conversion_buffer): New structure. (MAX_ALLOCA): New macro. (allocate_conversion_buffer): New macro. (extend_conversion_buffer, free_conversion_buffer): New functions. (ccl_coding_driver): Set coding->result. (decode_coding): Set coding->result to CODING_FINISH_NORMAL if this is the last block of source. (encode_coding): Likewise. Handle the source block as the last one only when the whole source text is consumed. (decode_coding_string): Handle the case that the output buffer is too small to decode the whole source text. Use allocate_conversion_buffer, extend_conversion_buffer and free_conversion_buffer, not get_conversion_buffer. (encode_coding_string): Likewise. (init_coding): Function deleted. (init_coding_once): Delete code to initialize conversion_buffer_size.

This is a list of known problems to date with the Windows NT/95 port of
GNU Emacs.

* Handle wildcards in ls-lisp (e.g., C-x d *.c).

* Interactive subprocess output is buffered in jerky

* Presently, C:\foo\bar and C:/foo/bar bring up two buffers on the same file.
  Solve this by adding a switch that canonicalizes path separators
  (e.g., make them all / or all \)?

* call-process-region: Another tricky situation with binary and text modes.
  An example by [e-mail]: use crypt++ to load compressed
  data into a buffer, edit the buffer, save the data back out.  (Also
  need to propagate the "/C" switch change sent the shell in crypt++.el
  back to the author.)

* Dired uses ls-lisp, which reports all files as being owned by the
  current user.  Need to dig through the security descriptor to extract
  the owner of the file (and the group?) using LookupAccountSid.

* Integrate the build for NT into the GNU config process once
  a decent shell becomes freely available

* Integrate networking.

* Fix Win95 subprocesses.