Git refreshing

Hi Peter,

Le 01/03/2015 17:00, Peter Uhnák a écrit :

Hi,

I'm facing a problem with branches and refreshing the code. Currently I
handle (except for commits) everything from commandline, but the problem
is that when I switch branches I don't know how to easily refresh the
code. What I am doing now is running loading Baseline again, but that
feels odd.

Not so odd, in truth. It looks like a perfectly fine solution, which, maybe, could be added to the concept of baseline in a repository.

Git revolves around branches, and being able to quickly create them,
switch them, and merge them is important.

So is there currently any solution that could resolve this? Maybe I
could run some code that would simply load what is on the disk? If I
switch branches for java code, the editor automatically loads new
content without me needing to do any extra work; it would be nice to
have something similar.

Well, as you have noticed in Java, C or C++, switching branches does not call make automatically :)

In Pharo, loading code means compiling it, so what you describe may be a bit harder to do. However, if we consider, as above, that working in different branches is linked to a "project" concept, with a baseline describing this project and its dependencies, then:

a baseline branch switch could mean: unload the current baseline, its dependencies, followed by the switch branch, followed by a load of the baseline (in the new branch).

On the git side, this wouldn't be hard to do.

Another approach, a tad simpler, would be : switch branch, then all packages already loaded in the image and existing in the branch are loaded with the head of the new branch. This one would be easy to implement with just a button "switch branch" in the GitFileTree repository inspector.

    Well, as you have noticed in Java, C or C++, switching branches does
    not call make automatically :)

You can use git hooks for that.

Ok. Then I would suggest you take the code source of CLANG, LLVM, OpenOffice or of the Linux kernel, you add the git hooks to call make, and you admire how fast changing branches becomes ;)

Want to switch branches before committing? Can do.

I'm not a git expert, but I know enough about git (and git users:)) that each and everyone may have its opinion on the way git should be used, and, as such, I've decided on GitFileTree that I wouldn't claim to hold the way it should be used™.

I will have to think about it. I am used to my git workflow and now I am
kind of lost how to use it properly from Pharo; since I am cut off from
git. One idea that is crossing my mind is having synchronized filedumps
(filetree) in a folder and not worry about git or monticello inside
Pharo at all. (And when I change code outside it would magically change
also inside).

Please explore the idea; just try to focus on that:

At the same time, there is just me and Max (and Esteban as well, I think) working on getting that git support working, so don't bash us too much. It's hard enough to get it done (look at Max questions about libcgit) that we don't need too much additional pressure.

    Well, as you have noticed in Java, C or C++, switching branches does
    not call make automatically :)

You can use git hooks for that.

Ok. Then I would suggest you take the code source of CLANG, LLVM, OpenOffice or of the Linux kernel, you add the git hooks to call make, and you admire how fast changing branches becomes ;)

Sure; large project would take long regardless but comparatively small (~10kloc) project would get checkouted fast, also it doesn't just completely rewrite the filesystem, but only the actual changes (which is for my use cases most of the time in the order of tens of files)

 

Want to switch branches before committing? Can do.

How? :-o If I switch the branch from git (command line) then there is disparity. 

I'm not a git expert, but I know enough about git (and git users:)) that each and everyone may have its opinion on the way git should be used, and, as such, I've decided on GitFileTree that I wouldn't claim to hold the way it should be used™.

Yes, I need to think about myself a bit... I am used to certain workflow and I guess I was greedy with wanting everything.

 

I will have to think about it. I am used to my git workflow and now I am
kind of lost how to use it properly from Pharo; since I am cut off from
git. One idea that is crossing my mind is having synchronized filedumps
(filetree) in a folder and not worry about git or monticello inside
Pharo at all. (And when I change code outside it would magically change
also inside).

Please explore the idea; just try to focus on that:

Adding it my (endless) todolist, I would need to learn a bit (=lot) about monticello first and how these things actually work inside Pharo so it could be done, but just like everybody I have a lot on my plate. (And I have also tend to eat more than I can chew.)

At the same time, there is just me and Max (and Esteban as well, I think) working on getting that git support working, so don't bash us too much. It's hard enough to get it done (look at Max questions about libcgit) that we don't need too much additional pressure.

I am really sorry if my message sounded harsh, I in no way intended something like that. Sometimes I am more greedy and picky about software than I should be. I'll add some positive emoticons next time so it will be more happy. :)

I am grateful for the work you have done, after all I do like it and I am happy I was able to make the switch to git, despite me missing features; so please don't my ramblings as bashing, they were intended (albeit bit directly) only as questions/ideas.