This is the mail archive of the
crossgcc@sourceware.org
mailing list for the crossgcc project.
See the CrossGCC FAQ for lots
more information.
Re: crunchy ICE
- From: ng at piments dot com
- To: Crossgcc list <crossgcc at sourceware dot org>
- Date: Sat, 08 May 2010 19:16:21 +0200
- Subject: Re: crunchy ICE
- References: <4BDF8B7C.9080709@piments.com> <j2h56d259a01005040058yaeccfee1p6ce30b5e4d97c82b@mail.gmail.com>
On 05/04/10 09:58, Martin Guy wrote:
Am I making some careless mistake , or is ARM crunch still no ready to use?
Works perfectly. I built gnuplot to check whether you had found a bug
in my stuff and it's fine.
Now please stop insulting me and treating me like your servant.
You puonish me for having been generous and kind to you, and I regret it.
M
Hi all,
some progress but still not too good on maverick crunch.
I have applied the patches found here, in the order shown in series.txt.
http://martinwguy.co.uk/martin/crunch/
They all apply cleanly after letting ct-ng 1.7.0 get the source and
applies it's patchset. Clean, without any fuzz, and I have a newly built
maverick toolchain. This encouraged me to press ahead with a system rebuild.
Having rebuilt the entire system and booted by via nfs I tested the
major fp intensive program I'm using gnuplot. I get a seg fault on the
code and data that I have working on non maverick builds with the same
kernel.
Digging into it a bit I find gnuplot is producing some pretty odd maths:
gnuplot> print log(1)
-5.51218129652977e+178
gnuplot> print log(10)
-1.12247324422731e-113
gnuplot> print asin(1)
5.29980882362664e-315
gnuplot> print cos(0)
5.29980882362664e-315
gnuplot> print sin(0)
0.0
Martin says he built gnuplot but not that he tested it.
Has anyone else produced a build that will work on this dodgy Cirrus
silicon ? Martin seems to have run some pretty thorough tests so it's
odd I'm getting such spurious results.
Thanks for any input.
--
For unsubscribe information see http://sourceware.org/lists.html#faq