Compiler toolchain roadmap

[Warner Losh]

On Apr 6, 2014, at 4:16 AM, Jordan Hubbard <jkh at ixsystems.com> wrote:

> Now it’s 2014 and apparently we can’t have nice things in the tree because of MIPS?   Maybe I’m over-simplifying the argument, but even simplistically I would easily understand

First off, nobody every said we can’t have nice things in the tree because of MIPS. Where was that said? It can’t possibly be true because gcc supports blocks in the tree, so there’s no impediment. LLVM-based things? Show me the money and bring one to the table and we can talk, but even then there’s clang support for mips, so again that’s not a big deal.

As for numbers, perhaps you are right about mips, perhaps not. There’s a thriving community, the code isn’t holding the tree back, and things do get fixed there. Maybe not as well as our ARM community, but it still us. I hear a lot of FUD and chest pounding about how it is holding us back, but I’ve yet to see any real evidence of that proffered. Mips and powerpc are in the tree because Juniper needs/wants them, and has been contributing fixes to the tree over the years.

External toolchain is coming along nicely given the timelines for 11, which is where we committed to having it done and removing gcc/binutils from the tree. clang is good, but it isn’t quite ready for the binutils removal yet. If the time comes and it isn’t done, then we can talk about how mips is holding things up (assuming the clang mips stuff doesn’t go in). But until then, show me the concrete examples where there’s an actual problem.

Warner