NOTES   [plain text]


wishlist - Tue Nov  2 15:22:58 PST 1993

* bcopy -> memcpy & friends.
  ** done 12/18/93
  
* remove static buffers.
* replace list & node cache with recursive obstacks, (xmalloc,
  getnode, getlist)
* check all io functions for error return codes.  also check all
  system calls.
* error check mkdir.

---
Old notes...

* All sizing limits are gone.  The rest of these items were incidental
  in that effort.

* login name from history was duplicated.  taught existing routine to
  cache and use that instead.  Also add routines to cache uid, pid,
  etc.

* ign strings were never freed.  Now they are.

* there was a printf("... %s ...", cp) vs *cp bug in history.c.  Now
  fixed.

* The environment variables TMPDIR, HOME, and LOGNAME were not
  honored.  Now they are.

* extra line inserted by do_editor() is gone.  Then obviated.  Editor
  is now called exactly once per checkin.

* revised editor behaviour.  Never use /dev/tty.  If the editor
  session fails, we haven't yet done anything.  Therefor the user can
  safely rerun cvs and we should just fail.  Also use the editor for
  initial log messages on added files.  Also omit the confirmation
  when adding directories.  Adding directories will require an
  explicit "commit" step soon.  Make it possible to prevent null login
  messages using #define REQUIRE_LOG_MESSAGES

* prototypes for all callbacks.

* all callbacks get ref pointers.

* do_recursion/start_recursion now use recusion_frame's rather than a
  list of a lot of pointers and global variables.

* corrected types on status_dirproc().

* CONFIRM_DIRECTORY_ADDS

* re_comp was innappropriate in a few places.  I've eliminated it.

* FORCE_MESSAGE_ON_ADD

* So I built a regression test.  Let's call it a sanity check to be
  less ambitious.  It exposed that cvs is difficult to call from
  scripts.