Hi Alistair, > Le 22 août 2018 à 12:02, Alistair Grant <akgrant0...@gmail.com> a écrit : > > Hi Christophe, > > On Wed, 22 Aug 2018 at 09:36, Christophe Demarey > <christophe.dema...@inria.fr> wrote: >> >> ... >> >> As a workaround for now, we are thinking to ship the pharo.version file >> along with the image to avoid to run images with compatible VMs that we >> cannot ensure they work well with a given image. > > Part of the issue here is that Pharo 7 is still in development. My > understanding is that within the development cycle we assume that the > latest VM is being used (if I'm wrong, I'm sure someone will correct > me :-)). > > I don't use Pharo Launcher, so I'm not familiar with it, but it seems > to me that for the current development version (currently Pharo 7) the > launcher needs to ensure that it has the latest stable VM.
You’re right but the problem is that Pharo Launcher has no way to know what is the pharo version of an image before running it and evaluate a small expression. To achieve that, Pharo Launcher uses a VM that is, at least, compatible with the image format number. We do not have any guarantee that the expression evaluation of the VM will work fine for all images. For now, to detect the pharo version of Spur images, we use Pharo 61 VM. You can assume it will work well for 6. images. What if we use the 70 VM to determine pharo version of 6.* images? It will probably work better for Pharo 7 images (assuming we use the latest 70 VM) but then, no guarantee it will work for 61 images. The core of the problem is that we should not be forced to run an image to determine its pharo version. Nevertheless, I think it would be good to use the latest available VM for a given image format to determine its pharo version. It should work in most cases. Cheers, Christophe
