address@hidden: negative impact of process-adaptive-read-buffering=t on AIX]

[Richard Stallman]

Can someone please investigate this?

Should the default value of this variable be set according to the type
of system?

------- Start of forwarded message -------
X-Spam-Status: No, score=0.0 required=5.0 tests=UNPARSEABLE_RELAY 
        autolearn=failed version=3.1.0
Date: Wed, 14 Nov 2007 00:58:40 -0800 (PST)
From: Gilles Pion <address@hidden>
To: address@hidden
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: 
Subject: negative impact of process-adaptive-read-buffering=t on AIX


(using Emacs version: GNU Emacs 23.0.50.1 (powerpc-ibm-aix5.3.0.0, X
toolkit), compiled directly from source fetched from CVS)

Since I had to migrate from Solaris to AIX I have noticed *very* poor tty
output performance inside emacs subprocesses. (using
process-connection-type=nil worked but I *need* a tty in many situations) 
I've upgraded from emacs-22 to version 23 hoping better results without any
success.

Yesterday, grepping inside the sources I've noticed this option:
"process-adaptive-read-buffering", but since the default value was already
the "supposed to be optimal" one, I've made a test with the "unoptimized"
value (nil), and strangely, it solved all my problems!

For example, using command "ps -ef" inside a shell buffer, gives:

with process-adaptive-read-buffering = t
elapsed:        69 sec

with process-adaptive-read-buffering = nil
elapsed:        10 sec


Is the documentation of the parameter incorrect or maybe the AIX
implementation broken? 

- -----
- --
Gilles
- -- 
View this message in context: 
http://www.nabble.com/negative-impact-of-process-adaptive-read-buffering%3Dt-on-AIX-tf4803615.html#a13742752
Sent from the Emacs - Bugs mailing list archive at Nabble.com.
------- End of forwarded message -------