Re: [arm-gnu] Possible corruption of s20 register with 2009q1-203

[Mark Mitchell <mark@xxxxxxxxxxxxxxxx>]

Alan Young wrote:

> I have been trying to track down a memory corruption problem for a
> couple of months now, I believe that I have isolated it to a particular
> C function. 

If you believe that it's a compiler bug, then you will need to post the
preprocessed source code for the file (which you can obtain with
-save-temps), together with all command-line options, and an indication
of which function you think is miscompiled so that we can investigate it.

However, if it's fixed in our latest release, we won't take any further
action; we don't make new Lite Edition releases based on old versions.
(That is, however, something we do for subscription customers with
long-term support contracts.)

> http://hardwarebug.org/2010/01/15/arm-compiler-update/)

Do not believe everything you read.  We believe that Sourcery G++ 4.4 is
a very solid platform, and we know that many of our major silicon and
operating system customers are using it.  It does of course have bugs,
as do all versions of GCC.  The code-size issues you are seeing are
likely due to changes to the defaults to get faster code at the expense
of slightly bigger code; you can adjust those options as you like.

-- 
Mark Mitchell
CodeSourcery
mark@xxxxxxxxxxxxxxxx
(650) 331-3385 x713