We have a large application that runs under QNX4 on an embedded x86 board and I'm looking for a way to propose a replacement. The decision has already been made to not upgrade to the latest QNX.

What I'm thinking of is some GPL/BSD/MIT licensed OS that could be enhanced to allow our product to compile and run almost unmodified.

Specifically, I need independent processes (not just tasks), networking, a filesystem (ideally compatible with the QNX filesystem), interprocess message passing (via /dev/mqueue/) and a shell (no graphics). We don't strictly need a realtime OS, as long as performance is good. And we don't use QNX's signature TDP distributed architecture.

Any suggestions?

Well it seems qnx is a great choice for your needs.

Although if your open to trying something diferent. I would suggest LynxOS or NuttX. I think bench testing with your program in place will obviously be a good measure before going live. You may want to try a few systems before you find the running balance you are looking for.

As well, if programming your application in asm is an option. The footprint and speed of your app will be greatly improved. Further making the app more compatible with qnx.

I hope this is of some help.

Neqtan

I saw both LynxOS and NuttX in Wikipedia's list of RTOSes. But LynxOS looks like it's essentially Linux, which might be an option, but a big job. Nuttx doesn't look like it has enough support for our x86 processors.

Programming in asm? We're not just starting out. We have a big product with multiple processes (multiple separate QNX executables all running and passing each other messages) written in C++ with many, many lines of code already written, debugged and working in the field.

I have in mind something like Xenomai Linux, which calls itself a chameleon that implements the APIs of VxWorks, pSOS+, VRTX and uITRON. If QNX were already in that list, I'd jump on it.

Maybe Xenomai can be made to work, but I'm just asking if anyone knows of another solution--if anyone else has found a QNX migration strategy.

Wikipedia is good but I personaly steer away from it.

My reference to LynxOS: http://software.intel.com/en-us/arti...-qnx-or-linux/

My reference to NuttX: http://nuttx.sourceforge.net/

My thought on app migration to asm, was based on documentation that qnx is mostly written in assembly. I gathered that your app was not at the early stages of dev.. It was merely a suggestion, no matter how rediculous it sounds .

Have looked into RTAI

Neq

Ken,

I'm writing an article about Migrating from QNX.

I wrote QNX or Linux http://software.intel.com/en-us/arti...-qnx-or-linux/

Is it strictly cost that has you changing platforms? I found or was quoted these prices.

QNX – depends on components purchased – several hundred to over $10,000 per seat - USD$9695 per seat (one time) and $1842 per seat per year support.
Wind River Linux ($3,000 to $15,000 per seat per year in 2006)
MontaVista Linux (over $8,000 per developer per year)
LynuxWorks Inc. ($11, 500/dev/yr - 2005)
iMedia Linux (USD$33/dev seat current)
TimeSys Linux (USD$5495/dev seat -2012)

Have you seen Yocto? http://www.yoctoproject.org/

What I like about Yocto is that it is free, based and aligned with OpenEmbedded and heavily supported by Intel.

I have a few questions that I need to ask for the article:

Reason for dropping QNX?
OS and platforms being considered?
problem so far porting from QNX?
Biggest problem so far?

Thanks,

Tom
twburger@gmail.com