fun w/ git

.gitignore

#
# .gitignore - inhibit git from tracking files matching the below patterns
#
#


# for directory tracking, add empty .gitignore to example/.
!.gitignore
!.gitkeep


# file browser metadata/artifacts
.DS_Store
.directory


# alt/aux version control tracking artifacts
.svn


# alt/aux tooling artifacts
.vagrant


# ignore secret/personal stuffs (except example files)
*.gpg
*.asc
.secret.d/*/*
!*__example


# unnecesary autosaves/backups
*~
*\#*
auto-save-list/*
~$*
.swp
*.swp

# temp ao cache files
var/*/*
tmp/*



# compiled ao binary
*.pyc
*.elc
*.o
*.a
*.jar
*.exe
*.obj
*.dll
*.class

# temp files
*\#*
*~
*.log

# MSVS project files
**/Debug/
**/Release/
**/Backup/
**/elog/
*.v11.suo
*.sdf
*.opensdf
# intelisense db
*.ncb
*.idb
*.pdb
#*.unsuccessfulbuild
#*.lastbuildstate
#*.vcxproj
#*.vcxproj.filters
#*.tlog
#*.log
#*.intermediate.manifest
#*.exp
#*.sdf

# never ignore these
!**/Utilities/**
!**/.gitkeep

git-annex w gitolite

# init
git init .
git commit -m"init commit" --allow-empty
git push --set-upstream origin master:master
git annex init <device-label>
# ^^ checks out branch annex/direct/master
git annex trust origin
git annex merge
git annex sync

# nb, git-annex.exe spawns several git.exe process

# nb, git-annex assistant will spawn a background ssh.exe to your origin remote

# nb, .git/annex/url is the URL for the webapps, great for monitoring activity, (since assistant --foreground is a bit dull) whee!

# EM on login: [Windows Scripting Host]"Can not find script file "C:Program Files(x86)\Git\git-annex-autostart.vbs"

# EM on 'git annex trust origin'
"""
 Remote origin does not have git-annex installed; setting annex-ignore

  This could be a problem with the git-annex installation on the remote. Please make sure that git-annex-shell is available in PATH when you ssh into the remote. Once you have fixed the git-annex installation, run: git config remote.origin.annex-ignore false
git-annex: cannot determine uuid for origin
"""

# nb, use git annex vicfg to edit configs
# nb, schedule jobs w/ : git annex schedule <remote> <expr>
# nb, prune git-annex history : git annex forget
# nb, sync / merge ... push local changes to remote, pull remote changes to local (respectively)
# nb, other notable command/options: webapp, watch, mirror <path> , assistant, dead <remote> , unannex <path>, status

# to sensibly add contents to git /o git-annex, (ex. binary files larger than 400k)
find * -type f -size +400k -exec grep -ILhq '.' {} \; -exec git annex add {} \;
# -or-
git config --global annex.largefiles "largerthan=300kb and not (include=*.c or include=*.h)"
# then, add matching files to the annex
git annex --auto add
# git annex assistant will automatically add non-matching files to git
git annex assistant



#nb, mysgit doesn't include 'file' cmd, alternatively to distinguish bin from text files, check for a null-byte w/i the first 1k, this would only occur in a binary file
# - Q: how to do this on msysgit: 
#   - w/ head and grep (there's no dd), nope just grep
#   - A: grep '.' -ILhq <file> # opts: treat binary files as non-matches, only succeed on non-matches, quiet output

# nb, remotedaemon looks neat, it watches remotes for changes to pull

git-stdin.sh

#!/bin/bash -x

# be careful not to pollute stdin prior to catching it here
cmd_git="${cmd_git:-git}"
obj=$(cat - | ${cmd_git} hash-object -w --stdin)


# ENV-overridable variables
msg_usage="
mysqldump | env uri_file='msyql-dump.sql.bkup' branch="mysql_bkup" cmd_git='git --git-dir=${HOME}/.extrefs/sql_bkup.git --work-tree=/tmp' ~/bin/git-stdin.sh
"
uri_file="${uri_file:-unnamed_file}"
branch="${branch:-master}"
obj_mode="${obj_mode:-100644}"



### functions
switch_branch(){
    local tmp_branch="${1:-$branch}"

    #+NB: if symbolic-refs succeeds, reset index to HEAD else it's HEAD@{1}
    #+TODO: test that $tmp_branch is ok
    git symbolic-ref HEAD refs/heads/${tmp_branch} &>/dev/null
    [[ 0 -eq $? ]] && ${cmd_git} reset --mixed &>/dev/null
}


### main
switch_branch
${cmd_git} update-index --add --cacheinfo ${obj_mode} ${obj} "${uri_file:-default_name}"

parent=$(${cmd_git} rev-parse "${branch}^{commit}" 2>/dev/null)
#if [[ -n $parent ]] ;then unset parent ;fi

#+TODO: fix write-tree, '-m text' message not working
#message="backup commit"
#+NB: for write-tree, tree must come before options
${cmd_git} update-ref "refs/heads/${branch}" $(${cmd_git} commit-tree $(${cmd_git} write-tree) ${message:+ -m $message} ${parent:+ -p $parent}  ;)

gitconfig

#
# git config - http://kernel.org/pub/software/scm/git/docs/git-config.html
# don't run the whole thing, rather it's likely best to copy/paste/pluck targeted lines

# identity
#git config --global user.name "niklauz"
#git config --global user.email khaije1@gmail.com
#git config --global user.signingkey <gpg-key-id>

# github authinfo
#git config --global github.user khaije1
#git config --global github.token <see-github-profile> #secret!!

# aliases
git config --global alias.stash-unapply '! git stash show -p | git apply -R'
git config --global alias.unstage "! git reset HEAD --" #nb: run w/o file args will reset entire index to HEAD
git config --global alias.undo-commit "! git reset --soft HEAD^"
git config --global alias.ls   "!  test -x $(which git-annex) && git annex find --include '*' 2>/dev/null ; git ls-files -mustdock ; git status -sb "
git config --global alias.logm "! git log --graph --pretty --decorate --color --date=relative --tags  --submodule=log  --find-copies-harder --stat HEAD"  #--notes --cherry-mark 
git config --global alias.logl "! git log --graph --decorate --color --tags --oneline --all --cherry-mark -- "
git config --global alias.last "cat-file commit HEAD"
git config --global alias.st "! git status -sb && [[ $(which git-annex) ]] && git annex find '*'"
git config --global alias.br branch
git config --global alias.co checkout
git config --global alias.mt mergetool

# environment defaults
git config --global core.editor "emacs -nw"    #Changing your Editor
git config --global color.ui true    #Adding Color, See all color.* options in the git config docs
git config --global help.autocorrect 120   #Ignore Typos

# merge
git config --global merge.defaultToUpstream true
git config --global rerere.enabled 1     #activate the RE-use REcorded merge REsolution option
git config --global mergetool.keepBackup false #if false git mergetool will remove tempfiles after successful merge
git config --global merge.tool emerge


# credential helper/cache , http://git-scm.com/docs/git-credential-cache.html
#git config --global credential.helper cache                    # (default: 5mins)
#git config --global credential.helper 'cache --timeout=900'    # set timeout to 15 mins
git config --global credential.helper 'cache --timeout=10800'    # set timeout to 3hrs



# Other Config Options
#    There are also a number of interesting options for packing, gc-ing, merging, remotes, branches, http transport, 
#    diffs, paging, whitespace and more. If you want to tweak these, check out the git config docs.

# git-annex
git config --global annex.largefiles "(largerthan=400kb)and(not(mimetype=\"text/*\"))"
git config --global annex.addsmallfiles false

# working w/ large files
#git verify-pack -v .git/objects/pack/pack-*.idx | grep blob | sort -k3nr | head | while read s x b x; do git rev-list --all --objects | grep $s | awk '{print "'"$b"'",$0;}'; done

gitolite--git-annex--integration--testing--status

PROBLEMS
none

OUTCOMES
No problems encountered

OUTLOOK
 - Nothing major - FWIW git-annex is working toward a big release in upcoming mo/yrs. It doesn't look like what I've tested, (the basics), will change but there may be some new capabilities/interactions of potential interest to examine and account for.
 
  - I may wish to add to documentation on one side of another to ease setup. It's all as it should be, just possibly unanticipated by those unfamiliar with git-annex.
  
 - *NB* git-annex support in gitolite is stable in every case I've tested, however there is a significant case that I have been unable to test for one simple reason: I don't understand the situation well enough to test it. This untested case is described below.
    - Case Description
      - USER desires that git-annex have two remotes: a bup remote and a (gitolite) git remote
      - USER desires to move data from the bup repo, through git-annex, to gitolite
      - USER expects the above process to be lossless as described and in reverse, i.e to creating a new bup repo, perhaps using git-annex, using only data sourced from gitolite without loss of data or feature-capability
    - Case Analysis & Questions
      - bup introduces a meta-packfile, (i.e. it is not native to git)
      - git-annex evidently handles bup's non-standard-git files, is this communicated in a git-standard way to gitolite? If so, how?


 - Additionally, I have not tested git-annex support in gitolite in the case where the git-annex 'assistant' functionality is used.


STATUS
 - DONE test basic functionality with gitolite v3 core
 - DONE patch lines mentioned in 'PROBLEMS'
 - DONE test remaining functionality with git annex (refex specs, acl, get, copy, etc)
 - NEXT test different annex special remotes(there are many, are all expected to work?): bup, rsync, directory, etc.
   - though untested and thus unclear, if any of the above cases are determined to be non-compatible could it be deemed an acceptable functional constraint and not a "blocker"?
   - to better document any capabilities/limitations it's appropriate to wait for the outcomes of these final steps, if only to produce more accurate expectations and/or documentation. Ideally someone who understands this well could address the underlying technical principles in clarifying what should be reasonably possible and/or reasonably expected.

merging tmp -slash- feat -slash- pu child branches

merging branches, while controlling history, is easier than I realized w/ 'git merge --squash'

it's great for cleanly merging work from a testing branch into a parent branch

before I was 
(1) checking out a branch, then 
(2) checking out the path of another branch with 'git co -- .'  to get update the working-dir without moving HEAD then
(3) unstaging everything so that the index would return to the HEAD commit
from there I would 
(4) merge the working-tree into the index to satisfaction, committing as I go 
finishing with a 
(5) 'git reset && git clean -df'  to remove any changes that I still didn't want


Instead... I should have been using 'git merge --squash' 
It's the same basic process but with much more manageable actions.



(1) 'git merge --squash <commitish>' merges (w/o making a commit) the commitish's state in to the index and working tree, showing a --stat style summary 
(2) 'git save stash --keep-index --include-untracked' saves the changeset w/o removing it
(3) make commits by strategicly adding ('git add -p') from the index (or the stash)
(4) 'git reset && git clean -df && git stash drop' to cleanup when finished


so now I add hunks for what I want for each commit, and just whittle away from the stash as I want, even switching branches to distribute specific changes without having to redo the process... (also potentially useful for regression testing)

split-wd-gitdir-configs

If I do this often I like to setup aliases as such:

alias git.centralrepo01="git --git-dir=/srv/repo/centralrepo01 --work-tree=/srv/repo/centralrepo01"

This can obviously be extended to great effect. For example I use it to access/maintain remote content under a local repo with a different username:

alias git.backupscripts="sudo -u backup -s git --git-dir=$HOME/tmp/bkup/scripts --work-tree=/var/backups/.scripts"

on most systems these will tab-complete and allow for:
git<tab><tab>
: git git.
git.<tab><tab>
: git.centralrepo01 git.backupscripts
git.c<tab>
: git.centralrepo01
git.centralrepo01 status
: <will show status of wd vs HEAD>

when combined with other git config.alias commands this can be really useful!

I'm looking at ways to use this to backup databases directly, (w/o temp files).