On 03/06/2018 10:51 AM, Cyril Ferlicot D. wrote:
Cyril,
For development I tend to use Metacello locking instead of hard-wired
dependencies in the BaselineOf and that works very well --- it
completely avoids the need to edit a baseline for development purposes
and this approach works really well for me ... perhaps we can discuss
this in more detail in a separate thread or even private email? This all
seems to be more complicated for you guys than has been my experience
and of course you guys _appear_ to be completely ignoring the Metacello
locking feature so I'm curious why ...
Dale
Hi Dale,
As asked I start a new thread.
I can see some possible reasons I do not use Metacello locking feature.
1) I did not understand well enough what it was
2) We have a way to work different and mine might be incompatible with
the lock feature
3) I just did not see how it could help me
To know which one is the reason I'll present you a little what I
understood of the lock feature and how I usually work and you'll be able
to advise me.
For the lock feature:
If I understood correctly it is a feature allowing to lock a certain
version of a project in the Metacello registry. For example if I load
Seaside 3.1, lock it and then install a project having Seaside 3.2 as
dependency, the lock will block the update of Seaside.
Okay, this is how a lock for a ConfigurationOf works, but the effect of
a lock on a BaselineOf is that the FileTree directory where the
BaselineOf is loaded from becomes the "locked version" ... in other
words locking a BaselineOf is really a lock on a git directory and
whatever SHA you happen to have checked out in the location will be used
... and this means that you can explicitily checkout a particular tag in
the git repo and that is the "locked version"
Now how do I usually work:
Let's take the example of MaterialDesignLite that I maintain.
The master branch is a branch that should only contains stable MDL
versions. Ideally I would like for each commit of this branch to be
tagged with a semantic versionning tag.
The development happen in the development branch and new features are
merged into this branch before making it to master.
In general I prefer to have stricter dependencies in master than in
development. So the baseline in the development branch can be different
than in master.
For all my projects I try to have a Jenkins building Pharo images for at
least master and development, and potentially big feature branches
before they are merged. The Jenkins also test the images.
For development builds how do you specify/control the version of each of
you 70 dependent projects for your Pharo build ...
The use case for using locks would say that you have 70 cloned git
repositories that you are using for development purposes: i.e, you
anticipate the possibility that you will need to make active changes to
anyone of those 70 cloned git repositories as you do your development
and I assume that you are always operating on the head of the
development branch for each of these 70 cloned repositories ...
Then you would simply lock all 70 projects to the filetree directory for
the projects and when you decide to update to the latest development
branch of a particular project you do a git pull in that clone and your
next build will use the new version of that project ...
When you make a change from within the image and commit a new version in
one of the projects ... again the next build will use the new SHA and so
on ...
Once you've done a lock to a cloned git repo, you don't need to do
fiddle with the BaselineOf, you simply manipulate the version that is
present in the cloned git repo to change which SHA you want to use and
all builds will simple use the "current SHA" for each of projects ...
and since you've got active clones being used, it is simple to commit
from your current working image ...
I know that I've simplified some things here and I still don't see what
you do as an alternative to what I am suggesting with locks ... but
hopefully I've given you a bit clearer picture of what I imagine, you
could/should be doing?
Dale
