[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Yes, embedded systems do need multi-arch!



As an Emdebian newcomer (but long-time embedded Linux afficionado) who is looking at Emdebian to form the basis of a commercial embedded thin client O/S for my employer, ThinLinx Inc (thinlinx.com), I was annoyed to discover that Emdebian Grip (stable, at least - I see from your FAQ that jessie-grip may be a different story, but that doesn't help me right now) lacks Multi-Arch headers in its package metadata, thereby rendering multi-arch effectively unusable even though the package contents themselves are multi-arch-ready.

The Emdebian FAQ states that "it [was] not expected that many devices running Emdebian Grip would need to use Multi-Arch", a decision which I assume was based on reasoning along the lines of "multi-arch is only for development, so you don't need it on a production system, right?". If that was your thinking, I am here to assure you that, from a commercial embedded Linux standpoint, you were wrong.

We do need multi-arch at runtime. We need it because not all code that we use is open source, or even available to us in source form. Some industry partners and/or upstream software suppliers are slow to adopt newer ABIs. Case in point is Citrix Receiver, formerly Citrix ICA Client. It's currently only available for ARMEL, and we would really prefer to use ARMHF. It's not so very long since ICA Client was only available for Standard ABI, and they didn't have an EABI build! Another example from the x86 world, that many of you may recall, was the Adobe Flash Player browser plugin - Adobe failed to produce a 64-bit port of their 32-bit plugin for many years after fully 64-bit Linux systems became popular.

I am persisting with Emdebian Grip, because I think that it offers enormous flexibility at a relatively affordable resource cost, but for now I'm going to have to maintain parallel ARMEL and ARMHF builds (either that, or extract ARMEL packages outside of package management on my ARMHF build, which I am loath to do).

But please, next time you make a decision like this, consult the broader community first, at least if you entertain hope of attracting serious interest from industry!

P.S. For others hoping for an ARMHF Citrix Receiver, we have nagged them about this, and they have assured us that they are working on it!



Reply to: