Downloading "powerpc-rtems4.10 gcc" for a few hours on Fedora 12
ralf.corsepius at rtems.org
Fri Apr 30 03:27:42 UTC 2010
On 04/29/2010 08:51 PM, Peter Dufault wrote:
> On Apr 29, 2010, at 2:40 , Ralf Corsepius wrote:
>> No, that's avoiding a prominent and often hitting source of errors: VMs (esp. vmware etc.).
> "I recommend that for developing for RTEMS you always use physical hardware and never an emulator".
> "I recommend that for maintaining earlier versions of your RTEMS software you always keep a dedicated physical machine and not a VM installation".
> OK, I'm starting to be nasty. I'll back off.
> I've recommended to a client that they use VM installs both for development of current releases and to maintain a VM snapshot for minor maintenance on earlier releases. What do you recommend instead?
Well, you are overinterpreting what I wrote.
If one is sure that a VM works and that the guest OS works in a VM, then
there shouldn't be anything wrong in using a VM.
That is, to getting started, one should
1. make sure the VM works
2. make sure the guest OS works
If these prerequisites apply, "getting an RTEMS installation into the
VM" should be straight forward and pretty simple.
It these prerequisites do not apply, a direct Fedora install is one
Provided what you wrote (low bandwidth), I would not want to exclude you
are having issues with network access (NIC?) with Fedora in your VM or
with running Fedora on your VM in general.
> Typically they might not have a Fedora release installed for development.
They don't have to use Fedora. Any of the supported Linuxes (I.e.
Fedora, CentOS/RHEL, openSUSE) is supposed to "work out of the box".
> There should be an RTEMS environment guaranteed to work well in a VM.
Well, this is hardly possible, because that's not an RTEMS issue, but a
guess OS and a VM issue.
RTEMS is just a "application package suite" running inside of an OS.
That said, there are alternatives to using VMs, e.g.
- Running Linux from an external drive (e.g. usbstick, usb-harddrive).
I know people did so in the past, I haven't tried this myself in recent
times, but getting this to work should be possible and not be too much
work with Fedora 12 and openSUSE-11.2.
[This also is nice in cases when "long term support" is a requirement.]
- Running Linux inside of chroots. This is how I test "non-native" linux
distro packages (e.g. openSUSE-11.2 package on Fedora 12)
- Use a dedicated spare machine. RTEMS HW demands are fairly low, so
decicating any semi-decent spare machine as dedicated "RTEMS devel
machine" also is an option.
- Multi-boot configurations.
> I think that having a CURRENT environment work well in a VM should also be a requirement, but that last statement can be argued about.
> You're saying that VMs aren't mature enough for use in the application of an RTEMS build environment.
No, that's not what I am saying.
I am saying, not all guest OSes run on all VMs. Some do/some don't, some
do sometimes, some work unreliable, ...
May-be you could try a different Linux into your VM?
> To me that's one of the less demanding applications for a VM,
Exactly. RTEMS demands on a VM rsp. on HW in general are comparatively low.
> and one of the ones that is important for embedded development. Am I wrong?
No, not at all.
More information about the users