Re: [ECOS] Development environment under Windows without cygwin?

[Alex Schuilenburg <alexs@ecoscentric.com>]

On 2011-05-13 09:29, Manuel Borchers wrote:
> [...]
> Currently, we have a quite old arm-elf toolchain (4.0.3) on the windows
> boxes.
You did not say which version of eCos you were using, but if you are
still using arm-elf this would indicate you are using eCos version 2
which could be the first of your problems. Much has happened since eCos
2.0 was released. With the eCos 3.0 release in 2009, we upgraded the ARM
compiler to arm-eabi and gcc version 4.3.2.

I strongly suggest you switch to eCos 3.0/anoncvs and the arm-eabi gcc
4.3.2 (cygwin hosted) we contributed to the eCos 3.0 public release


>  So grabbed a 4.0.4 compiler for my linux box. Using a virtual
> machine with windows I was quite successful with this setup. Linking
> went fine in this scenario, but executing the small demo app failed.
> After a little research I found out what was failing: the DHCP managment
> thread bailed out with an assertion that the stack base wasn't properly
> aligned. Using arm-elf-nm I saw that the assertion was quite right. The
> linker linked the stack to an odd address! I double checked the ld file,
> and there clearly was the ALIGN(0x4) at the .bss section. So, I guess I
> found a linker/compiler bug. I could work around the problem by adding
> an align attribute to the stack definition, but I've a little
> stomach-ache doing so and that such an error might occure elsewhere,
> too.
You probably have found a linker/compiler bug.  Not many people
appreciate the effort we put into testing the GNU tools on real hardware
and fixing some or other rare edge condition or the like before we ship
them to customers or contribute them. I certainly don't think you can
take it for granted that any toolchain you download will just work -
just read this list's archives...


> And now back to the original topic:
> I tried to get the ecoscentric-provided arm-eabi toolchain running under
> windows. I failed doing so. Even after installing cygwin (which wouldn't
> be that easy to install to the lab PCs) and shuffling dll's around. I
> guess, the toolchain is too old to get it properly running on modern
> cygwin/windows machines.
I disagree and I find this surprising.  The 4.3.2 arm-eabi toolchain
contribution is only 2 years old and we have had no issues using it with
a current cygwin on XP or 7, although there have been problems with
cygwin on vista but I believe the cygwin team have sorted out those
issues quite a while ago. For eCos/eCosPro 3.0 we only officially
supported Windows XP, although we do have customers successfully using
the tools on Windows 7.

Our switch to native Windows for the GNU tools was mainly for the
speedup and to remove the complexities some of our customers were
encountering with installing and administering cygwin on their
tightly-locked-down corporate desktops.  If you have full control of the
PC, you should not encounter any difficulties (although from what you
mention, you dont have that luxury).


> So the question is: are there plans by ecoscetric to contribute a newer
> arm-eabi windows toolchain to the public? That would be really great,
> because I really don't like to build one on my own for windows... That
> one in the commercial version being cygwin-free raised my attention and
> this question ;)
Sorry, I cannot comment on any plans regarding contributing the newer
4.4.5 toolchain or our eCosPro 3.1 release for commercial reasons.

However, there is no reason I know of why you cannot get the cygwin
hosted 4.3.2 arm-eabi toolchain we contributed 2 years ago working for
you. There are many Windows users on this list that are successfully
using it without any problems.  Perhaps you could be specific about what
kind of issues you are seeing so one of them could answer :-)

Cheers
-- Alex


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss