Date: Tue, 10 Sep 1996 13:51:30 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: DJ Delorie <dj AT delorie DOT com>
Cc: djgpp-workers AT delorie DOT com
Subject: Re: V2 alpha 980908 -- GDB 4.16
In-Reply-To: <199609090348.XAA28882@delorie.com>
Message-Id: <Pine.SUN.3.91.960910133902.10696G-100000@is>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

First, thanks to everybody who worked on this.  Comments:

	1) The info files are identical to what I got with 4.12.  Didn't 
the docs change at all?  Seems incredible with all the versions that 
passed below the bridge in between.

	2) Gdb seems to be quite a lot faster at startup, but still 
painfully slow when the first breakpoint in a given module is hit.  I 
recall somebody saying that GDB (or is it Binutils?) think they cannot 
rely on `fseek' with DJGPP, and that might be the reason for the slow 
speed.  If that indeed is the problem, I think the new buffered I/O code 
solves it, so we might as well tell Binutils to cheer up and hum away.  
(Sorry, I didn't have time to download the sources and look into them.)

	3) If you exit GDB before the debuggee finishes, you are asked to
confirm with a `y' or `n'.  So far so good, but if the debuggee switched
stdin to BINARY mode, GDB hangs: you can only reboot.  That looks
precisely like the bug with printing long screenfuls that I corrected in
the previous version (and which works correctly in this port): when you
call `fgetc' (or any other buffered I/O function), you should momentarily
switch stdin to O_TEXT. 

	4) If I terminate a long backtrace display with a `q RET', I get 
a message telling to "expect SIGINT when resumed".  Huh?