Man page: new material in BUGS section.

1 files changed, 19 insertions, 2 deletions

diff --git a/pw.1 b/pw.1
index c09bd2c..b90a34b 100644
--- a/pw.1
+++ b/pw.1
@@ -371,6 +371,8 @@ Unexpected conditions like out of memory result in abnormal termination (abort).
 This was written over the course of a couple of hours, and tested only
 interactively.
 
+Many of the issues which follow are easy; patches welcome.
+
 The program doesn't respond to window size changes.  The display format, such
 as the handling of control characters, is hard-coded.
 
@@ -381,13 +383,28 @@ terminal abstraction library/data.
 The intervals cannot be dynamically adjusted.
 
 There is no support for unwrapping long lines, which would be useful for
-copy and paste.
+copy and paste. However, the features like
+.B :w
+to save the displayed data somewhat compensate for this.
 
 During the
 .B :!
 command's execution, the TTY settings are not restored.
 
-Most of these issues are easy; patches welcome.
+The timeout isn't applied to the arrival of data into the FIFO, but on
+the arrival of data into the program. Therefore, the grep stack feature doesn't
+necessarily do what you want in terms of the timeout-based refresh behavior.
+Consider a situation in which ten lines are arriving per second at a steady
+rate. The default 1s timeout therefore never goes off. The long
+refresh interval is what is updating the display, every 10s by default. Now
+suppose you put in a
+.BI :g pattern
+command which only passes every 20th line. Effectively, it now looks as if
+a line appears every 2s, and so the 1s timeout should be going off to
+refresh the display. Unfortunately, that's not how it works. The program
+is still receiving ten lines per second from standard input, and that's where
+the timeout is applied.  Until that is fixed, the way to get more timely
+refresh behavior under heavy filtering is to play with the long interval.
 
 .SH AUTHOR
 Kaz Kylheku <kaz@kylheku.com>