HI, I finally found the solution to our problem that prevented svn access for i18n committers. There is only one last issue that we need to fix ( but this will have to wait after Easter as I am going see my family ), but it doesn't affect everybody.
So basically, people who are committers for their team should be able to connect with svn+ssh to svn.mageia.org. Testing access -------------- To check, that's rather easy , type in a shell of any UNIX : ssh svn.mageia.org It will ask your password ( the one of transifex , bugzilla, etc ), and if the password is correct ( and the login too ), it will show you a error message and tell "you are a member of theses group : " + list of group. Something like this : Last login: Thu Apr 21 22:52:31 2011 from nanardon.zarb.org You tried to run a interactive shell. Sorry, you are not allowed to execute that command. You are member of the following groups : mga-users mga-sysadmin mga-packagers mga-ml_moderators mga-packagers-committers mga-board mga-founders mga-i18n mga-i18n-committers Connection to svn.mageia.org closed. ( the exact message may reworded later to show more information, but basically, that's normal to have it on svn ). Doing a checkout ----------------- First, you need to know where is the software. For example, let's take drakpxelinux : https://transifex.mageia.org/projects/p/drakpxelinux/ This one is rather easy, using the web interface, we can find it : http://svnweb.mageia.org/soft/drakpxelinux/ ( that's not always easy, of course ). Next, all you need is to do a checkout, with svn co. The url for the repository is svn+ssh://svn.mageia.org/svn/soft/ So in our case : svn co svn+ssh://svn.mageia.org/svn/soft/drakpxelinux/trunk drakpxelinux Then that's the part where you commit the .po ( I am not translator so I do not know the detail ) and then you commit with "svn commit", and likely notify your team. what if it doesn't work. ------------------------ If you have a message like this : Can't open file '/svn/soft/db/txn-current-lock': Permission denied then you suffer from the problem I spoke at the beggining. It is easy to check, when you do the test for the acces when you do the check with ssh and the message do not say that you are in mga-shell_access group, then you are suffering from the problem I need to fix, and this will not work. If it say that you are in this group, then this should work. If it doesn't accept your password, check that it is the good one, or check that the user is the proper one. YOu can use -l option for ssh, or the configuration file, or [email protected] syntax. If it still doesn't work, try to find a admin ( bugzilla, irc, email ), we will look at it. Of course, if you are not the committer for your team, it will not work. If you forgot your password --------------------------- Since all committers have been granted a restricted shell account, the password can no longer be reset without a admin intervention. So if you lost the password, the link on identity will not work. Same stuff, try to find a admin for that. What to improve ---------------- * First, typing the password each time is not fun. If like me, your password is as long as the first chapter of Harry Potter, this become quickly annoying. You can ask to a sysadmin to place a ssh keys if you want, just contact us ( better open a bug so we do not forget ). Something to automate this is planned since months ( we just need to have a slightly simpler interface on identity, perl coder help welcome, and to test the script to export keys, and another script to import existing keys from the disk ). * second, the whole process of matching from tx project name to svn url is tedious. Oliver proposed something back when we discussed at FOSDEM but I forgot the exact idea ( but he can surely explain ). But we can surely make it better, now that tx-client work fine. Any proposal is welcome. * third, there is not technically enforced restrictions, mainly because that's more complex. And we didn't decide if we should add them ( I posted on -dev on this subject, your input is welcome ). People with committers access can commit everywhere in the svn so far, but obviously, we would prefer that no one wreck havoc, even if this can be reverted ( depending on the type of havoc as committing a whole dvd is gonna make admins very sad ). * four, the notification system is quite basic, there is 1 big ml for the whole repository. Maybe having a better dispatch system would help. To finish, the process is not set in stone, and given that I forgot what we discussed with Oliver and Wobo at FOSDEM, it may be different from what we discussed so any comment would be welcome. -- Michael Scherer
