Differences
This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revisionNext revision | Previous revisionNext revisionBoth sides next revision | ||
git-migration [2009/04/19 21:12] – nick | git-migration [2009/05/09 02:46] – kelnos | ||
---|---|---|---|
Line 5: | Line 5: | ||
===== Resources ===== | ===== Resources ===== | ||
+ | * [[http:// | ||
* [[http:// | * [[http:// | ||
* [[http:// | * [[http:// | ||
- | * possible conversion tool: [[http:// | + | * <del>possible conversion tool: [[http:// |
- | * another possible conversion tool(?): [[http:// | + | * <del>another possible conversion tool(?): [[http:// |
* [[http:// | * [[http:// | ||
+ | * stuff about [[http:// | ||
+ | * Behdad' | ||
- | ===== Missing | + | ===== Stuff That's Done ===== |
- | * pre-commit permissions-handling hooks | ||
* svn username -> realname+email mapping | * svn username -> realname+email mapping | ||
- | * We have one of these for the git-svn repos, but people might not want the email I picked to be the one used. | + | * People with a https account on svn.xfce.org: |
+ | * [[http:// | ||
+ | * SVN -> git conversion script | ||
+ | * See [[http:// | ||
+ | * Generated | ||
+ | * See [[http:// | ||
+ | ===== Stuff In Progress ===== | ||
+ | |||
+ | |||
+ | |||
+ | ===== Missing Stuff ===== | ||
+ | |||
+ | * pre-commit permissions-handling hooks (Brian has some ideas about this) | ||
+ | * commit mail hook (Jannis will take care of this) | ||
+ | * I18n stats pulled from git (Stephan, I presume, will do this) | ||
===== Open Questions (please discuss!) ===== | ===== Open Questions (please discuss!) ===== | ||
* The big one: do we really want to do this? Are there any legitimate objections or reasons why we shouldn' | * The big one: do we really want to do this? Are there any legitimate objections or reasons why we shouldn' | ||
- | * I obviously think this is the right way to go, but maybe there are convincing reasons to the contrary. | + | * I obviously think this is the right way to go, but maybe there are convincing reasons to the contrary. |
+ | * I prefer a DVCS over CVCS, since it allows me to use VC features when I do not have access to an inet connection. I don't have a particular preference for git, prolly any DVCS will do. --- // | ||
* Note that hg, bzr, mtn, etc. aren't really on the table. | * Note that hg, bzr, mtn, etc. aren't really on the table. | ||
* Do we continue to have separate trees for ' | * Do we continue to have separate trees for ' | ||
Line 26: | Line 43: | ||
* I'd also prefer one big tree. With SVN we just had two repositories (xfce, goodies) but with git each component will have its own repository anyway. It's just a matter of which URL to use to check them out then. --- // | * I'd also prefer one big tree. With SVN we just had two repositories (xfce, goodies) but with git each component will have its own repository anyway. It's just a matter of which URL to use to check them out then. --- // | ||
* I'm ok with that too, as long as we can still separate the permissions. How are the po permissions handled in git, is it possible to set this on directory level? --- // | * I'm ok with that too, as long as we can still separate the permissions. How are the po permissions handled in git, is it possible to set this on directory level? --- // | ||
- | | + | * Git has no support for access control. If we want fine-grained permissions we'll have to implement an access control script in one of the hooks that can reject a commit. --- // |
+ | * I prefer one big tree too, I don't see a good reason why we should split that up. --- // | ||
+ | | ||
* Possibly we'd do this with a [[http:// | * Possibly we'd do this with a [[http:// | ||
* I would prefer the freedesktop way for this. Xfce core components start with xfce/ | * I would prefer the freedesktop way for this. Xfce core components start with xfce/ | ||
+ | * Branch/tag name rewriting: we have old nasty branch/tag names like " | ||
+ | * Current [[http:// | ||
+ | * The GNOME guys rewrote their commit messages on the fly while converting to make them more like the git recommended commit msg pattern (max 72 char single line shortlog, blank line, freeform larger description). | ||
+ | * Probably not. Our commit messages are all over the place with regard to how they' |