You totally rock! I'm so eager to have iceberg on 64bits. It is the last block for me to move my projects into the 21st century ;)
thanks, Norbert > Am 19.06.2017 um 08:00 schrieb esteba...@gmail.com: > > Hello! > > This is my weekly ChangeLog, from 12 June 2017 to 18 June 2017. > You can see it in a better format by going here: > http://log.smallworks.eu/web/search?from=12/6/2017&to=18/6/2017 > > ChangeLog > ========= > > 15 June 2017: > ------------- > > * I finished an iteration for allowing cherry-pick into > [iceberg](http://github.com/pharo-vcs/iceberg). > > Now, I just need to make it work on 64bits and we will be ready-to-go for > releasing it :) > > > 13 June 2017: > ------------- > > * Still "iceberging", I fixed an annoying bug when browsing diff with > parent (in all screens that have it), > who was trowing a DNU because of course, first commits do not have a > parent to compare with. > > * I've been working on iceberg 0.5. It will incorporate some needed > features (to make it appropriate to > work on pharo itself, and better for users). > > This was also my work last week, but I forget to announce it here ;) > > So far, I added following features: > > === Better diff handling > It was very naive because it was comparing always all present packages > which does not scales on > large repositories. Now, I implemented a double dispatch that works in > different scenarios: > > * commit -> commit: It relies in libgit2 to take the list of changed > packages. > * commit -> branch: it takes last commit on branch on continues as > commit->commit scenario > * loaded code -> commit: this is the most complicated scenario. It > compares what is in image with what is in commit to take a correct diff. > > With this, now diffs are performant in all cases (bah, if you have a > commit that changes > 300 packages, you will suffer anyway, but I would argue that is not a very > good codign practice) > > Also with this you will finally be able to add not-dirty packages to your > repo and commit them :) > > === Branch sync with disk copy > I had a discussion with author of Iceberg (Nico) on this: originally he > weanted to treat loaded code in > the image as a got working-copy itself, so he was taking the information > of the branch and keepong it. > > Problem with this is, if you change the repository branch (in command line > or another image) and then > go back to your current image, you do not have any feedback... and that is > severely confusing. > > So now I'm just taking the information from git repository. That means you > can have differences between > what you loaded and your current branch, but that's, IMO, perfectly > reasonable (you can always switch back or > merge or whatever you need to do) > > === Other > * Minor cleanups > * Other bugfixes > > > cheers! > Esteban