X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
Message-ID: <41012e39$0$172$cc7c7865@news.luth.se>
From: Martin Str|mberg <ams AT speedy DOT ludd DOT ltu DOT se>
Subject: Re: GDB on target w/o coprocessor
Newsgroups: comp.os.msdos.djgpp
References: <pqRLc.37726$5D1 DOT 1638448 AT news4 DOT tin DOT it> <2ma6n0FkuqsjU1 AT uni-berlin DOT de>
User-Agent: tin/1.4.6-20020816 ("Aerials") (UNIX) (NetBSD/1.6Q (alpha))
Date: 23 Jul 2004 15:26:49 GMT
Lines: 26
NNTP-Posting-Host: speedy.ludd.ltu.se
X-Trace: 1090596409 news.luth.se 172 130.240.16.13
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de> wrote:
: If memory serves, there's a fundamental limitation that effectively
: makes it impossible to use GDB on an FPU-less box.  You may have better
: luck installing an FPU, or doing your debugging on "bigger" hardware.

No that's not necessary. I've had this problem, and it seems that gdb,
the program or the math emulation endlessly loop if the math emulation
needs to execute.

The solution is to tell gdb not to stop on math traps/exceptions
(unsure of the term). I think I did this by saying "sig SIGEMT
nostop" to gdb. Please read the documentation as I'm unsure of the
exact command I used. Perhaps it was necessary to say "sig SIGFPE
nostop" too. Or perhaps only the second one. It was some time ago. If
you search the mail archives you'll find my postings but I think it
might have been on the djggp-workers list.


It might also be worth to know that some of gcc's C++ code generation
has been for 486+ a very long time as it uses some 486 instruction. I
don't know if it's been corrected in later versions.


Right,

						MartinS