[SOLVED] n00b needs some git hacking

Weapon S · 09-25-2012, 01:53 AM

I've made a terrible mistake. I did:

Code:

$git init
$git add -A *.cpp
//I didn't want all the sub-directories as well...
$git rm -r *.*
//Hm, doesn't work. Do --force or --cached?
$git rm -r --force *.*

WHOOPSIE!
It seems all the objects do exist in the git directory. They're just totally unindexed. I'd like to access them. Anybody have any experience with this? Know a good resource? HALP! Please.
(The fun part is that before a commit there is no HEAD and usual revert techniques don't work.)

konsolebox · 09-25-2012, 02:12 AM

When did you make the commit? Is it possible for git to have objects when it didn't make a commit yet?

Weapon S · 09-25-2012, 10:51 AM

THERE IS NO COMMIT!
References used:
http://git-scm.com/book/en/Git-Internals-Git-Objects
http://linuxconfig.org/Bash_scripting_Tutorial

Code:

find .git/objects -type f  > /media/BAD/SDevX/my/PE_gitbu/recover/index.txt
#edit file to contain only last directory (without /) and filenames
#write bash script script1
chmod +x ./script1.sh
./script1.sh edited_index.txt

The bash script:

Code:

#!/bin/bash
# Link filedescriptor 10 with stdin
exec 10<&0
# stdin replaced with a file supplied as a first argument
exec < $1
while read LINE; do
    echo `git cat-file -p ${LINE} > /media/BAD/SDevX/my/PE_gitbu/recover/try1/${LINE}.cpp`
done
# restore stdin from filedescriptor 10
# and close filedescriptor 10
exec 0<&10 10<&-

Now I have a directory full of oddly named files. The names are based on the contents only; i.e. not file location. Tree objects describe the (original) location of files. None of these objects was a tree

Well, at least I have my code back.

konsolebox · 09-25-2012, 11:13 AM

Well what just confuses me is why would there be even a record if no commit was made. 'git init' just initializes the repo, and 'git add' just adds references -> would that have to include data among files since by commit it has to update itself again anyway. Why would it make a cache just in case. So what's confusing is why would you even have a data about those files. Perhaps it's just a necessary step to protect data - not really as a strict compliance to necessary procedures.

Weapon S · 09-25-2012, 03:33 PM

I was as surprised by the fact that git can remove files I hadn't added to any tree, as I was that git actually keeps a copy of deleted files.