Open CPI installation material still points to CentOS 7, which goes End of Life soon.
Is there a plan to migrate to a different version of Linux as the default target environment?
Alternatively, what problems are my team likely to encounter using Fedora Core, if any?
We are considering Rocky8 as a replacement for CentOS7. This would be a good topic for discussion. Will this work for everyone? Is there a better alternative?
There shouldn’t be any major problems creating an RCC platform to support Fedora Core as development host. Only one that comes to mind is potential issues with any FPGA vendor tools.
In my mind there is only really one reason why CentOS 7 is still suggested as the main distro:
- Xilinx 19.2 is the current recommended toolset, and it segfaults in various ways on any OS other than CentOS 7 (most of, if not all of, these segfaults involve XSim).
Most of the current devboards use this for their kernels, and the E310 image is also Xilinx Petalinux 19.2.
Noting that Xilinx 21.2 causes issues with the kernel driver atm, that is not a suitable replacement. Neither is anything newer.
Issue 3520 demonstrated that Xilinx 20.1 and Xilinx 20.2 at least avoid the XSim segfaults in newer Ubuntu versions, but you’d need to make new embedded RCC Platforms for them (although this is probably copy and paste).
I think you could also just move to Ubuntu 22.04 now and keep Xilinx 19.2 as long as you have a different simulator to use (I cannot assure that I am correct here, although testing this would only take a few hours).
With regards to what the next default target environment becomes, I’d personally want it to be Ubuntu 22.04 or an alternative distro of similar age.
Is there demand for a very long LTS? If so, then you probably have to pick something other than Ubuntu.
Be warned that with newer OSes you increase the risk of hitting kernel changes, although I recently added fixes up to Kernel 6.5 (not exhaustively tested tho).
Hi, we were expecting alignment to Ubuntu LTS releases (and the V2.4.7 roadmap issued in September 23 seemed to say as much). The idea that a slightly left-field choice will be used is a bit disconcerting.
Ubuntu is a headline supported OS for the Xilinx tools and seems an obvious choice; we would not want to have an OS that is not well-supported or well-used with Vivado as it adds risk which is a problem with fixed-time or fixed-budget projects. It is certainly a lot easier to introduce new engineers to OS platforms that are well used when it comes to them finding material to orientate themselves.
To alleviate some of your concern, I don’t think anyone is suggesting removing Ubuntu support, or considering it a substantially lesser development platform.
CentOS 7 is considered the current “primary” development platform for a few reasons:
- History.
- It was the most important OS for most users for a long time.
- It “works” best.
- It is the most tested OS.
- A new developer is likely to hit the least issues, none of which should be completely blocking.
- It’s stable and aligns to RHEL.
- I think this is slightly more of a legacy concern?
- Some organisations like the really long LTS times of RHEL.
Looking at the newest Xilinx Vivado Supported Operating Systems list, it seems Xilinx still haven’t selected a non-enterprise CentOS 7 replacement which is somewhat baffling.
Given we want to remove CentOS 7, and still align with a free OS explicitly supported by Vivado, Ubuntu is the only choice atm (noting that Vivado 2024.1 is actually dropping Ubuntu 18 and 20).