For those who’ve been interested in what went on during the last two weeks, where I visited Lille, I’ve compiled a short report. Regarding my work on Git I’ll continue to report on the status from time to time.
I want to thank everyone at RMoD once again for the warm welcome and the support I received. Special thanks to Ben and Camillo for providing a place to crash at. Thanks also go to all of you guys and girls on the list for pushing Pharo constantly and for showing interest in my work. It’s very encouraging. Cheers, Max Git: - Esteban Lorenzano and me worked together on preparing the libgit2 and libssh2 libraries for integration into the VM - Igor Stasenko worked with me to solve a couple of problems I had with NativeBoost. Especially callbacks can be tricky - Learnt how to build the VM in debugging mode so that I could debug FFI calls in XCode - Stefan Marr worked with me on moving the tests from Phexample to SUnit - The implementation now enables writing of blobs, trees and commits, cloning of remote repositories (https), fetching from remote repositories (https), pushing to remote repositories (https) - Authentication with remote repositories via SSH is working but clone, fetch, push don't work yet (problems with the libssh2 interaction that I wasn't able to resolve yet) - I set up an initial build infrastructure on the INRIA CI server (https://ci.inria.fr/pharo-contribution/job/LibGit2/) - We defined a rough roadmap for what needs to be done in the near future: 1. finish the low level libgit2 abstraction (offer a minimal API that hides the bindings; we don't want users to use the bindings directly) 2. we already have a prototype of a FileSystem wrapper for Git. We want to use that on top of the libgit2 abstraction layer 3. we already have a Monticello FileSystem wrapper prototype. We want to use that to abstract from the actual storage method. Together with the Git-FileSystem wrapper this should make it very easy to continue (for now) using Monticello and the existing GUI tools while using Git as a backend for storage. - On Friday 21 I gave a short demo at RMoD on the work accomplished and what the plans are for the future Fuel: - Martín Dias and I worked together on: - debugging a problem with large object graphs - preparing a new baseline for Fuel 2 - moving the benchmark suite to SMark - setting up a benchmark build on the INRIA CI server which will help us track performance changes when introducing changes - defining a rough roadmap for Fuel - Had a discussion with Usman Bhatti about the uses and the future of Fuel in Moose
