Hi Hilaire, Moving this to pharo-dev... On Thu, Mar 08, 2018 at 11:34:55AM +0100, Hilaire wrote:

I give again a try to bootstrap (after my last summer fail). Here how far I went:

Do people really like git or pretend to because everyone is using it.

Maybe there's a better source management system out there, but for me git is much better than any of the alternatives I've used (rcs, svn, vcs and a couple I've forgotten (cms?)).

1. Fetched the repo https://github.com/pharo-project/pharo. I have to fetch a fresh one as updating my repo failed with the git command pull/fetch/merge/reset hard does not work. Incredible how simple thinks does not work simply with git.

I consider myself far from a git expert, so take the following with a grain of salt... You haven't given much detail, but when you have a clone think of the branches that are upstream as being "read only" (typically master and development). They should only ever be modified by merging in the branch from upstream. If you're writing (committing) to your clone, it should always be in a separate branch. When updating, e.g. master, from upstream, get in to the habit of enforcing fast-foward only e.g.: git checkout master git fetch --prune upstream git merge --ff-only upstream/master git push It's more steps, but will ensure you don't get in to the problem you describe above. git pull will do a merge if necessary, making future updates hard / impossible.

2. Followed instructions to build a bootstrap.

BUILD_NUMBER=42 BOOTSTRAP_ARCH=32 sh ./bootstrap/scripts/bootstrap.sh

The script I've been using is: #!/bin/bash export JOB=bootstrap export BOOTSTRAP_ARCH=32 export PHARO_VERSION=60 export BUILD_NUMBER=$(date "+%m%d") ./bootstrap/scripts/bootstrap.sh It would be good to have a readme in the bootstrap directory that provides basic instructions on what the key variables are and how to run the bootstrap.

The resulting image bootstrap-cache/bootstrap.image is built after 1 or 2 hours. Its size is 3,4MB, when started there is only blank screen, I guess it is the expected outcome.

My 4 year old Dell XPS-13 (i7) takes about 30-40 minutes to do the build.

3. Followed instructions to load pharo environment:

??BUILD_NUMBER=42 BOOTSTRAP_ARCH=32 ./bootstrap/scripts/build.sh

This is called by bootstrap.sh, so it shouldn't be necessary for you to call it directly.

Several attempts were needed and the build.sh script is fragile and fail when at the repeated attempts, some cleaning were needed.

I suggest two things :

1. you add -p to your mkdir commands (in build.sh and get-vm.sh scripts) as it make your script less fragile ! 2. Also checking for the existence of the Pharo.source will avoid to have duplicted uncessary downloaded source file version. (Pharo.source1, Pharo.source2):

if test ! -e PharoV60.sources ???? then ???????????? wget http://files.pharo.org/sources/PharoV60.sources fi

Now the image is building, will see next.

I agree that it would be nice if it was possible to avoid downloading the files each time, and also possible separate the process in to a number of stages, and restart from a nominated stage. Maybe the stages could be: 1. Download all requirements (stable vm (6.0), target vm (7.0), bootstrapImage.zip and anything else I've forgotten). 2. The build steps in bootstrap.sh. 3. The build steps in build.sh There could be a "clean" option that would be called as part of the CI process to ensure that everything is always up to date. Esteban and co.: How open are you to refactoring the scripts to support this? Or is there something I don't understand that means it doesn't make sense? Cheers, Alistair