Error on new installation

I have been a long time user of git for little home jobs. My version of git (except in the very early days) been obtained as a standard debian binary package.

I have been developing some software on my home machine for a web site I run at a hosting company. I have a "site" branch for the version that is configured for running at the site, but up until now I have been manually transfering the files to the site I thought I had updated in the last commit.

I have now realised that if I stick a git repository on this hosting computer where my web site application has to be, I could git push my site branch to it on update and then use a post receive hook there to checkout the newly received version - thus automatically keeping the right files in sync.

Unfortunately this hosting company does not have git installed, so I downloaded the tar.gz version of 1.6.0, unpacked it and from inside the new directory did a make install.

This appears to have put key programs in ~/bin and the remainer in ~/libexec/git-core

As a first test, I went into the directory where one of my applications resides and did

git init git add . git commit

I then tried to clone back this repository with the following results

(NOTE mb is stanza in my ~/.ssh/config file which fully defines host/username/identity etc for my server at the hosting company)

alan@kanger:~/dev$ git clone --upload-pack=/home/melindas/bin/git-upload-pack ssh://mb/home/melindas/public_html/football/ Initialized empty Git repository in /home/alan/dev/football/.git/ error: git-upload-pack: git-pack-objects died with error. fatal: git-upload-pack: aborting due to possible repository corruption on the remote side. remote: fatal: exec pack-objects failed. remote: aborting due to possible repository corruption on the remote side. fatal: early EOF fatal: index-pack failed alan@kanger:~/dev$

the question is - why does git-pack-objects die. I presume the repository is not corrupt, it has been newly created. Does it rely on some other programs on the host that might not be there?

I use sshfs (my hosting don't permits Shell Acess) to mount my "sftp" server on my filesystem and it works welll (a litle bit slow, but works).

sshfs -o workaround=rename user@host:/home/user/  ~/web/hosting/

With this command I can just do:

cd ~/web/hosting/my_project/ git clone ~/projects/my_project git-checkout -b production #performs some server specific modifications git-checkout master git-merge production

#work offline iin my project on ~/projects/my_project/

cd ~/web/hosting/my_project/ git pull # and my project is up-to-date

fusermount -u ~/web/hosting/ # to unmount

from Brazil
                            Felipe

On Tuesday 02 Sep 2008, Felipe Carvalho Oliveira wrote:

Interesting - but I changed my mind in the end and use a post-commit/post-merge hook to see if I am now on my "site" branch on my home change and then rsync the working tree over to the server. It means I don't need the whole repository on the server, and I also can use git-describe --tags to give me a version no reference file