From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <058401c03e91$28be15e0$1d08ff87@oemcomputer>
From: "rob pike" <rob@plan9.bell-labs.com>
To: <9fans@cse.psu.edu>
References: <200010250640.e9P6eGG56159@ducky.net>
Subject: Re: [9fans] bug in rio
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Date: Wed, 25 Oct 2000 10:26:30 -0400
Topicbox-Message-UUID: 1d906e8a-eac9-11e9-9e20-41e7f4b1d025

Yup.  This is a known bug in rio and is a side effect of the property that all
graphics requests from clients are not mediated by rio but go directly
to the display, unlike in the predecessor system, 8 1/2. I thought about
putting some sort of "grab" protocol in the system to prevent the screen
mess but decided it's rare enough not to merit it.  The situation doesn't
make me happy, but the overall design does.  The red marks are a reminder
that I've never been able to get it right.

-rob

----- Original Message -----
From: Mike Haertel <mike@ducky.net>
To: <9fans@cse.psu.edu>
Sent: Wednesday, October 25, 2000 2:40 AM
Subject: [9fans] bug in rio


> to reproduce:
>
> 1.  create a window and enable automatic scrolling on the button 2 menu.
> call this Window 1.
> start a process in that window that will continuously print output, e.g.:
>
> while (~ x x) {
> echo hi
> sleep 1
> }
>
> 2.  move another window (call it Window 2) slowly across the front of
> the scrolling window.  the borders of Window 2 will be "captured" by
> the block copies due to the scrolling in Window 1.  you will see
> garbage copies of the border slowly scroll off the top of Window 1.
>
> i don't know any way to make the garbage go away except by hiding
> and then unhiding Window 1.  unlike other window systems i have used,
> rio does not seem to have a "refresh" command.  i imagine this is because
> reliable backing store ought to make that unnecessary.  alas...
>
> (N.B. the lack of a refresh command is also annoying in the case
> that kernel messages pollute your display; i occasionally get stray
> interrupt messages that seem to somehow be associated with the Tekram
> scsi controller.)
>
>