Sounds awesome. Getting the basics rock solid and as a first class server citizen. Sweet! Phil On Thu, Feb 7, 2019, 16:09 Esteban Lorenzano <estebanlm@gmail.com wrote:
Hi list,
We started Pharo 8.0 development and we wanted to share (and discuss, if needed) what is our current Roadmap for Pharo 8.0. As you can see, Windows is getting some love, and also UI. Anyway, here it is:
Image === 1) Missing parts for headless VM to work (explained a bit later) 2) We need to improve Epicea speed. And in general, source access speed. We want to remove the old changes file (since Epicea already does that works and a lot more). 3) Improve Refactors 4) Improve Calypso 5) Introduce Spec2 (our re-work on this framework). - We also want to migrate our tools to it (Inspector, Debugger, Spotter and Calypso are the remaining parts). We will see how much of this migration can be done.
VM/Low-level side === 1) headless vm We want to have a real headless VM and make it our default VM. To it, most of the work vm-side is already made by Ronie, but there are missing parts: - a build on windows - image side capabilities: we use SDL2 to start the world, and it mostly works... but not completely.
One cool thing of this is that we will -finally- be able to clean the event handling, which is ugly (and works bad).
2) Windows several missing/non working parts: - file primitives are slow. This is because they rely in old APIs and we need to put them in "state of the art". - libgit2 does not processes long paths. We workarounded the problem with tonel, but at a point we need to take care about this. Real problem with this is we need to contribute the solution to libgit2, but this is also good Open Source policy (contributing back). - OSSubprocess in windows. We believe we need to extend OSSubprocess (our solution to communicate with system) to windows. And we believe is possible ;)
3) ThreadedFFI. It is already too much time since we have this in agenda. Is time to make it real.
4) memory policies. Tweaking the VM to enhance its memory usage is possible, but hard. We want to adopt an scheme of "memory policies" that will allow users to pick what they need.
Process === 1) We will add multiple source directories to Iceberg. This is needed to allow us to put all Pharo sub-projects into an unique project without breaking modularisations.
Others === 1) Launcher - Launcher us getting a new UI - Tests - It needs to be more solid (in part, that's the reason why we want OSSubprocess in windows). 2) Cargo - We need to revisit cargo (a new dependency manager) and at a point decide if it will fly or not :)
Nice to have (most probably not this version, but in our TODO) : - embedded VM - event driven VM - what happens if we split VM into main thread and vm thread?