This is one of the most pathetically complex things I've ever had to work with. It CONSTANTLY gets in the way of actually getting work done - I spend more time f***ing around with pushes and adds and branches and checkouts that I do actually making code changes!!! What a P.O.S.
This diagram[^] near the bottom of the post says it all.
I think you've just saved me at least a week with that post. I'm looking for a place to host a project and for a VCS. The stuff I was reading recommended GIT. I won't be following it up any further. I've used SVN and its survivable as a team member but I dread having to manage it.
"The secret of happiness is freedom, and the secret of freedom, courage."
Thucydides (B.C. 460-400)
Sounds more like your workflow is the problem to me. I tend to script any repeated tasks with more than a couple steps, so my workflow generally becomes: get latest code -> make changes -> run script to make commits, pushes, etc. which is basically the same as using anything else.
Personally, I wish we used Git or Mercurial here, because our commit process strongly discourages putting partially complete things in the repository, and I'd like to have some of the benefits of version control (especially revert) when I'm working on something larger that has be made as one commit to the repository. I wonder if there's a way to set up Git locally to push to CVS...
which involves checkouts, fetches, and pulls, none of which make any sense to me in my angry state. I want to simply check out the latest version. How hard should that be?
run script to make commits, pushes, etc.
The fact that you have SCRIPTS to do those things shows how complex they are. Commit to staging area. Push to remote. Rebase, pull, checkout, WTF??? Again, a simple "commit the changes to the repository" would be sufficient, and again, because I'm so angry that every time I try to start some productive work I end up first spending an hour (YES, A F***ING HOUR) fighting Git, I really have no patience for listening to how "it's my workflow."
Maybe tomorrow I'll be in a better mood.
Care to share your scripts? I might learn something!
Generally I don't need more than a pull to get the code, then commit and push to submit it. I've had a few projects with branches, but I wasn't switching frequently. Maybe I just haven't worked in a large enough project. I'm not sure where my scripts are at the moment though, I fear they may be on my now dead laptop, because that's where I've done most of my Git based work from, but most of them were either just a series of commands I performed routinely (e.g. go to root of git repo, commit all changes, push) with some parameters for branch names, comments, etc. Nothing extremely complicated.
I only had a few more complex ones, like one that went back and branched a project I had gotten way ahead on for a class at each commit with a message containing a chapter number (fortunately, I was basically using a template for my commit messages) so I could just checkout the code for that chapter and submit it for grading, and a second to merge back some file adds when I realized I had forgot to add some files to some of the earlier commits rendering many of the older branches uncompilable, but those were one-time scripts that were faster to write than to do manually 20+ times. The second one could probably be modified into a decent merge script if I could find it.
Marc Clifton wrote:
I really have no patience for listening to how "it's my workflow."
I probably could have worded that a bit better, I don't necessarily mean your personal workflow, but also the workflow that's been imposed upon you (similar to the reason why I can't make commits when I would prefer). I don't find myself changing branches often, but our VCS is so messed up that I have to keep a separate local workspace for each branch because switching between them somehow renders the workspace un-buildable, and even simple tasks like merging a change to another branch becomes a real pain. If I had to work in an environment where I was frequently branching and merging on this system I feel that I'd be in the same position, but not at the fault of the VCS system, the problem would be the poorly managed repository (to be honest, no one actually knows how to make it build, everyone just copies a working copy from someone else, checking it out from the repo doesn't work and no one has ever bother to figure out why and fix it...I suspect the reason is buried somewhere in the thousands of build errors that are just ignored).
Yeah, each step requires multiple commands; I'm not sure if I should like that.
My typical workflow, when working on the main branch:
hacka hacka hacka
because that's part of our job, too.
Sort the mess I coded into easily digestable commits. UI sucks, but does the job.
(at command line, this would be git add, git commit)
alternatively / if this is not good enough:
git rebase -i pub/master
to reorder, clean up and combine commits.
git fetch pub
to get all changes from the other guys.
git rebase pub/master
Put my pending changes on top of the other guys changes
Look at the commits the other guys made, just to get an idea what they are on.
git push peterchen master
backup current state to remote repo (often involving -f)
Build the code. If it compiles, it's good to go!
git push pub master
Push my changes to the main repo
Sometimes, curse at whoever made a push pub master in the meantime
repeat the recent steps since fetch, omitting all the quality checks to get the code pushed and pray that I didn't break the pub repo
The data model takes a bit to wrap your head around. git stores revisions, but most commands actually move around changes between revisions. Branches and tags act like "pins", everything that is not directly or indirectly pinned gets garbage collected.
I have a local "temp" branch that I use for intermediates I want to save before messing around. (you could use git stash, but it's only good if you are 101% sure it will be short-lived).
Larger independent developments go on a branch. master is merged into that dev branch from time to time, when the code is good to go, dev branch is merged into master.
This just shows that your problem isn't really with git, it's with you lack of knowledge of it. I'd imagine you're kind of a green horn to the technology, and you haven't quite figured out how all of these things actually benefit you, if you take a moment to learn them.
The fact that you even have "repeated tasks with more than a couple steps" is WTF-worthy in something which is supposed to make your life easier. In SVN you have one-click (or one-command) 'update' and 'commit' for the most common tasks. If you can't do that in GIT without scripting then it is objectively worse for the 99% case.
I meant that as a general statement, not as Git specific. I script anything and everything I can. For example, since I find myself working with Base64 encoded files often, I now have scripts on my path to encode and decode Base64, a set for strings and a set for files (I could combine them, but that would be more work than having 2 sets honestly). As far as Git goes, the only common task I've encountered with more than one step is commit+push, but not every commit needs a push.
There are Git/Mercurial to SVN bridge tools; you could run one of hte former locally and just push your changes to SVN once you have it finished.
Did you ever see history portrayed as an old man with a wise brow and pulseless heart, waging all things in the balance of reason?
Is not rather the genius of history like an eternal, imploring maiden, full of fire, with a burning heart and flaming soul, humanly warm and humanly beautiful?
Training a telescope on one’s own belly button will only reveal lint. You like that? You go right on staring at it. I prefer looking at galaxies.
-- Sarah Hoyt
Last Visit: 31-Dec-99 18:00 Last Update: 21-Sep-14 17:07