| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
|
|
| |
This casues the SQL parsing to skip these files. Thus we have to
handle the branches separately and be careful to exclude trunk
while doing so.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This database will be used to provide redirection in weblinks etc.
Note: I had to abandon the while look with git-foreach-ref as it meant the
global variable was out of scope due to piping.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
The git-svn import stage doesn't automatically handle tags as well as it could.
In order to make tags show up properly on master, we inspect each tag and
try and find the first commit where it actually changes. We can then tag
it via an annotated tag. Much nicer.
|
| |
|
|
|
|
|
| |
Passing --author and --date to the commit command works fine for *author* commit dates
but not the actual *committer* dates (and name).
This ensures that the values are correct.
|
| |
|
|
| |
This prevents newer Mandriva tags and branches leaking into Mageia GIT repos
|
| |
|
|
| |
You can only skip commits that are *after* the last squashed commit.
|
| | |
|
| |
|
|
| |
This will confuse our final push at the end, so we want rid of these
|
| |
|
|
|
| |
This deals better with the case when there are old stale remotes
in the git repo (from the initial Mandriva svn conversion)
|
| |
|
|
|
|
|
|
| |
This is mostly useful during testing, but if we have to do several
runs to ensure the data is 100% correct (i.e. picking the correct
initial svn revision to import and any subsequent commits to squash)
then reusing the data collected here (perhaps with manual log editing)
will be very useful.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
As some sources were imported gradually in different commits,
it's neater to squash those commits together, even if this
means doing it out of order.
This causes less file churn overall and prevents the whole
delete and readd problem (breaking git blame and friends).
Sadly due to the use of binary data in the svn commits
which 'patch' cannot apply, we have to use svn merging
to ensure things are done properly. This means creating
a dummy svn repo and synthesising commits before exporting
it to form the initial squashed commit.
|
| | |
|
|
|
This script helps to automate a task that is, in a large part, manual.
It takes the git-svn repostitory from the Mandrvia import and guides
you through the task of adapting it and importing the Mageia SVN history
on top.
You first have to pick the trunk revision at which you want to reset the history.
This should be the point in time when Mageia imported the sources.
It will then help you find the correct Mandriva SVN revision to use as your
starting point. It will then synthesize a commit to reset the repository
at the same state as the Mageia subversion at the chosen revision. This
should result in minimal file churn (i.e. it's should be mostly
s/Mandriva/Magiea/ and a few other tweaks). In an ideal world it would contain
no other fixes but in order to minimise bogus git-blame results it should
avoid big 'delete-then-add' commits.
Once this is done, the script then synthesises the internal git-svn metadata
and tricks git-svn into continuing the import.
It will likely break if any branches/tags have the same name in Mageia
as existed in Mandriva. Fingers crossed that doesn't happen :)
|
