# newbie with broken workflow 6 messages from 2008-04-03 to 2008-04-04. Participants: James Kingston, Santi Béjar, Daniel Barkalow. Thread: https://gitlist.dev/t/12981 ## James Kingston, 2008-04-03 21:45 Subject: newbie with broken workflow Message-ID: <9d27c6320804031445p3fc80115n59218673dd587795@mail.gmail.com> URL: https://gitlist.dev/e/9d27c6320804031445p3fc80115n59218673dd587795%40mail.gmail.com ``` I am a few days into git and have settled into a workflow that is apparently very broken, and I'd appreciate it if someone with a more experienced eye can point out where I went wrong. The official SCM tool at my work is ClearCase Remote Client, and it's vile... pessimistic locking, huge bureaucracy just to create a branch (I love experimental branches), no licenses for the native client. My plan was to use git (Git-1.5.4-preview20080202.exe) to track my clearcase view directory, and push changes into that repo when convenient from my working directory repo (clearcase makes non-checkedout files read only, which makes working directly in the view directory Ugly). Here is a timeline of how I tried to do this: /c> cd /c/cc /c/cc> git init /c/cc> git add * /c/cc> git commit /c/cc> mkdir /c/wd /c/cc> cd /c/wd /c/wd> git clone /c/cc /c/wd> cd cc/ /c/wd/cc> ls /c/wd/cc> vi config.ini /c/wd/cc> git checkout -b config_changes /c/wd/cc> git diff /c/wd/cc> git add config.ini /c/wd/cc> git commit /c/wd/cc> git checkout master /c/wd/cc> git merge config_changes /c/wd/cc> vi config.ini /c/wd/cc> git push /c/cc /c/wd/cc> cd /c/cc /c/cc> vi config.ini # doesn't show the change made in experimental branch, though gitk shows the patch /c/cc> git status # says that config.ini has been modified /c/cc> git add config.ini # grasping as straws /c/cc> git commit # changes nothing, as far as I can tell /c/cc> git reset --hard #still changes nothing. My diffs are there in the log, just out of reach I have much to learn, but in the short term I just need to get those changes into clearcase so QA can pull them ``` ## Santi Béjar, 2008-04-03 22:03 Subject: Re: newbie with broken workflow Message-ID: <8aa486160804031503j1ff3adbcu83f41eef14347370@mail.gmail.com> URL: https://gitlist.dev/e/8aa486160804031503j1ff3adbcu83f41eef14347370%40mail.gmail.com In-Reply-To: <9d27c6320804031445p3fc80115n59218673dd587795@mail.gmail.com> ``` On Thu, Apr 3, 2008 at 11:45 PM, James Kingston wrote: [...] > > /c> cd /c/cc > /c/cc> git init > /c/cc> git add * > /c/cc> git commit > /c/cc> mkdir /c/wd > /c/cc> cd /c/wd > /c/wd> git clone /c/cc > /c/wd> cd cc/ > /c/wd/cc> ls > /c/wd/cc> vi config.ini > /c/wd/cc> git checkout -b config_changes > /c/wd/cc> git diff > /c/wd/cc> git add config.ini > /c/wd/cc> git commit > /c/wd/cc> git checkout master > /c/wd/cc> git merge config_changes > /c/wd/cc> vi config.ini > /c/wd/cc> git push /c/cc git push updates the branches, not the working copy, so: cd /c/cc git pull /c/wd/cc HEAD > /c/wd/cc> cd /c/cc > /c/cc> vi config.ini # doesn't show the change made in experimental > branch, though gitk shows the patch > /c/cc> git status # says that config.ini has been modified > /c/cc> git add config.ini # grasping as straws > /c/cc> git commit # changes nothing, as far as I can tell > /c/cc> git reset --hard #still changes nothing. My diffs are there > in the log, just out of reach > Santi ``` ## James Kingston, 2008-04-03 23:41 Subject: Re: newbie with broken workflow Message-ID: <9d27c6320804031641s2a1578fcmd7271a06842803fe@mail.gmail.com> URL: https://gitlist.dev/e/9d27c6320804031641s2a1578fcmd7271a06842803fe%40mail.gmail.com In-Reply-To: <8aa486160804031503j1ff3adbcu83f41eef14347370@mail.gmail.com> ``` On Thu, Apr 3, 2008 at 5:03 PM, Santi Béjar wrote: > git push updates the branches, not the working copy, so: > > cd /c/cc > git pull /c/wd/cc HEAD > Santi > Thank you. I also found the FAQ in the wiki, so the reasons are clear. ``` ## Daniel Barkalow, 2008-04-04 00:05 Subject: Re: newbie with broken workflow Message-ID: URL: https://gitlist.dev/e/alpine.LNX.1.00.0804031957010.19665%40iabervon.org In-Reply-To: <9d27c6320804031445p3fc80115n59218673dd587795@mail.gmail.com> ``` On Thu, 3 Apr 2008, James Kingston wrote: > I am a few days into git and have settled into a workflow that is > apparently very broken, and I'd appreciate it if someone with a more > experienced eye can point out where I went wrong. > > The official SCM tool at my work is ClearCase Remote Client, and it's > vile... pessimistic locking, huge bureaucracy just to create a branch > (I love experimental branches), no licenses for the native client. My > plan was to use git (Git-1.5.4-preview20080202.exe) to track my > clearcase view directory, and push changes into that repo when > convenient from my working directory repo (clearcase makes > non-checkedout files read only, which makes working directly in the > view directory Ugly). > > Here is a timeline of how I tried to do this: > > /c> cd /c/cc > /c/cc> git init > /c/cc> git add * > /c/cc> git commit > /c/cc> mkdir /c/wd > /c/cc> cd /c/wd > /c/wd> git clone /c/cc > /c/wd> cd cc/ > /c/wd/cc> ls > /c/wd/cc> vi config.ini > /c/wd/cc> git checkout -b config_changes > /c/wd/cc> git diff > /c/wd/cc> git add config.ini > /c/wd/cc> git commit > /c/wd/cc> git checkout master > /c/wd/cc> git merge config_changes > /c/wd/cc> vi config.ini > /c/wd/cc> git push /c/cc > /c/wd/cc> cd /c/cc > /c/cc> vi config.ini # doesn't show the change made in experimental > branch, though gitk shows the patch > /c/cc> git status # says that config.ini has been modified > /c/cc> git add config.ini # grasping as straws > /c/cc> git commit # changes nothing, as far as I can tell > /c/cc> git reset --hard #still changes nothing. My diffs are there > in the log, just out of reach > > I have much to learn, but in the short term I just need to get those > changes into clearcase so QA can pull them So you should actually do: /c/wd/cc> cd /c/cc /c/cc> git pull ../wd/cc But wouldn't that fail, because the files are read-only? I think you'd do: /c/cc> git fetch ../wd/cc /c/cc> git diff --name-only HEAD FETCH_HEAD Tell ClearCase you want to edit those files /c/cc> git merge FETCH_HEAD Commit your ClearCase stuff Also, IIRC, the sort of normal workflow for dealing with foreign SCMs that need local working directories is actually to put them in .git/info/something, and use git diff to get the changes in each of your git commits and git apply (using it like "patch", but smarter) to modify the working directory. -Daniel *This .sig left intentionally blank* ``` ## James Kingston, 2008-04-04 14:08 Subject: Re: newbie with broken workflow Message-ID: <9d27c6320804040708l6eace184q9677ca75729b4eef@mail.gmail.com> URL: https://gitlist.dev/e/9d27c6320804040708l6eace184q9677ca75729b4eef%40mail.gmail.com In-Reply-To: ``` On Thu, Apr 3, 2008 at 7:05 PM, Daniel Barkalow wrote: > /c/wd/cc> cd /c/cc > /c/cc> git pull ../wd/cc > > But wouldn't that fail, because the files are read-only? Actually, I just tried and it worked beautifully. The directory is writable, so git replaces rather than modifies the files. Then I can use CCRC to search for hijacked files, check them out in one swoop with the option to keep my modifications, then check them all in as a group. ``` ## Daniel Barkalow, 2008-04-04 16:30 Subject: Re: newbie with broken workflow Message-ID: URL: https://gitlist.dev/e/alpine.LNX.1.00.0804041228250.19665%40iabervon.org In-Reply-To: <9d27c6320804040708l6eace184q9677ca75729b4eef@mail.gmail.com> ``` On Fri, 4 Apr 2008, James Kingston wrote: > On Thu, Apr 3, 2008 at 7:05 PM, Daniel Barkalow wrote: > > /c/wd/cc> cd /c/cc > > /c/cc> git pull ../wd/cc > > > > But wouldn't that fail, because the files are read-only? > > Actually, I just tried and it worked beautifully. The directory is > writable, so git replaces rather than modifies the files. Then I can > use CCRC to search for hijacked files, check them out in one swoop > with the option to keep my modifications, then check them all in as a > group. Ah, yes, I guess that would work. Although I suspect that it would be annoying to recover if you failed to get the locks for any of them, since you're missing CC's expected path for conflict resolution. -Daniel *This .sig left intentionally blank* ```