On Sat, Aug 15, 2009 at 03:01:02PM +0100, Roger Leigh wrote: > On Sat, Aug 15, 2009 at 03:42:00PM +0200, Guido Günther wrote: > > Hi Roger, > > thanks a lot for the explantion! The plan is to do it quiet similar to > > what you describe but using git-fast-import instead. > > > > On Sat, Aug 15, 2009 at 02:08:09PM +0100, Roger Leigh wrote: > > [..snip..] > > > As an example, take what's currently done to import an orig source > > > (approximately) > > > > > > - check if working tree is dirty > > > - switch to orig branch > > > - unpack source > > > - remove and add files to make the working tree match the branch > > > - commit > > > - checkout master branch > > > - merge > > Have a look at git-import-orig --fast-import. This is how > > git-import-orig handles this (quiet similar to what you describe below) > > and how git-import-dsc(s) will handle all this in the future. You can > > give the ancestors via "from" and "ancestor". This is very similar to > > how git-commit-tree works except that you don't even need a temporary > > index. > [...] > > See above. Using git-fast-import has the advantage that we don't even > > need to unpack the source explicitly. We can feed it right from tar. > > It's just that I didn't get around to do it for git-import-dscs(s) as > > well. (https://honk.sigxcpu.org/piki/projects/git-buildpackage/) has > > this as TODO item (more git-fast-import support). > > This sounds great. Using git-fast-import should be even better for > importing .orig.tar.gz. For the DSC, don't you still need to > unpack in order to apply the .diff.gz though? In which case, can > you still use git-fast-import, or would using my suggested method > be required? I can't think of a way to skip the unpack for the dsc case so suggestions on this would be welcome. We can still use git-fast-import thouh (so we can use the same code paths). Cheers, -- Guido
-- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
