bug#727: compilation-disable-input

[Stefan Monnier]

>> I'm not sure what this is intended to do.  Sending just one EOF  
>> character to a process creates one EOF if many conditions are met.   
>> But a second read from the same pty will continue to look for input.
>> 
>> The following code:
>> (when compilation-disable-input
>> (condition-case nil
>> (process-send-eof proc)
>> ;; The process may have exited already.
>> (error nil)))
>> 
>> Causes my grep buffer to start with a ^D (the EOF character that  
>> process-send-eof sends).  This goofs up the parsing of the lines.

> I can't reproduce this (I don't see any ^D character in grep buffers).
> Could you send a recipe for reproducing this bug, as well as your system
> specs?

I remember seeing exactly these kinds of ^D under Mac OS X where
process-send-eof causes them (either for pipes or for ttys, can't
remember which).

I even added a workaround for this problem in pcvs.el's
cvs-parse-process:

  (when (eq system-type 'darwin)
    ;; Fixup the ^D^H^H inserted at beginning of buffer sometimes on MacOSX
    ;; because of the call to `process-send-eof'.
    (save-excursion
      (goto-char (point-min))
      (while (re-search-forward "^\\^D+" nil t)
        (let ((inhibit-read-only t))
          (delete-region (match-beginning 0) (match-end 0))))))

Not sure if it's related.


        Stefan