Solomon, Well done, Mr. Someone! Our toolchain has indeed been a big source of headaches for all involved.
A couple things: 1) Why stick to the 4.x.x series? More generally, is it worth the potential trouble to bump the major version? I'd be interested to measure performance gains with a more recent compiler (maybe in the less prominent codecs which haven't been completely hand-rolled in asm). 2) Testing. This change has the potential to introduce all sorts of insidious bugs. I'm usually of the "break-then-fix" mentality, but I feel like we should get some degree of testing done across targets/configurations before you merge this. It looks to me like you've eliminated all the gory patches we used to rely on, too. Great work! Franklin ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐ On Friday, March 27, 2020 3:06 PM, Solomon Peachy via rockbox-dev <rockbox-dev@cool.haxx.se> wrote: > I'd like to bump these toolchains to match what we're using on all MIPS > and the various hosted targets. I brought this up a while back and the > general concensus was "after the 3.15 release". Which is now past. :) > > I built a couple of m68k and arm targets cleanly, but I only have access > to one, namely the clip+. > > Are there any objections to my doing this? We will of course need the > various builders to update their toolchains too. The good news is that > 4.9.4 toolchains build cleanly even with the latest gcc 10, so that's > one major developer experience wart that goes away. > > I also tried to bump the sh toolchain to 4.9.4, and it ran out of space > on the target, and I didn't feel like trying to debug that, especially > as I haven't owned an archos target for well over a decade. Rather than > spending that effort, IMO it makes more sense to finally bite the bullet > and formally retire the Archose targets, and all HWCODEC stuff while > we're at it. > > Thoughts, comments, objections? > > - Solomon > -- > Solomon Peachy pizza at shaftnet dot org > High Springs, FL ^^ (email/xmpp) ^^ > Quidquid latine dictum sit, altum videtur. >
