I recently took over package maintenance for Audacity. So the first thing I want to do is clone their development tree, create a branch from the latest release and star tracking patches against it as needed. The Audacity team uses Subversion, while I prefer Git, so I attempted to use Git to clone their tree.
The problem is that it's been HOURS and Git hasn't finished cloning the tree.
The steps I took are:
1. mkdir Audacity; cd Audacity 2. git svn init --prefix=origin/ -b branches -t tags -T trunk http://audacity.googlecode.com/svn/audacity-src 3. git svn fetch --log-window-size 10000
It's the third step which is taking forever to complete. I left it to run overnight (10 hours) and it never finished. I paused it, came to work, resumed it and it's still going.
Should this process really take this long? Or is there a faster, more efficient way to do this? I know that part of what'll make it take so long is the amount of history that has to be extracted from Subversion. But is that many hours normal or did I do something wrong?
Thanks in advance.
On Wed, 8 Jan 2014 08:27:42 -0500 "Darryl L. Pierce" mcpierce@gmail.com wrote:
I recently took over package maintenance for Audacity. So the first thing I want to do is clone their development tree, create a branch from the latest release and star tracking patches against it as needed. The Audacity team uses Subversion, while I prefer Git, so I attempted to use Git to clone their tree.
The problem is that it's been HOURS and Git hasn't finished cloning the tree.
The steps I took are:
- mkdir Audacity; cd Audacity
- git svn init --prefix=origin/ -b branches -t tags -T trunk
http://audacity.googlecode.com/svn/audacity-src 3. git svn fetch --log-window-size 10000
It's the third step which is taking forever to complete. I left it to run overnight (10 hours) and it never finished. I paused it, came to work, resumed it and it's still going.
Should this process really take this long? Or is there a faster, more efficient way to do this? I know that part of what'll make it take so long is the amount of history that has to be extracted from Subversion. But is that many hours normal or did I do something wrong?
in my experience it can take hours to clone large SVN repos (similar to 13k revisions in audacity), I usually use git svn clone --stdlayout <svn_url> but it should be equivalent to your procedure
somewhere under .git you should see a file that changes during the fetch phase (IIRC it lists the grabbed revisions), so you will know it does something, also if svn:// url works, it should be faster than http://
Dan
On Wed, Jan 08, 2014 at 02:38:14PM +0100, Dan Horák wrote:
in my experience it can take hours to clone large SVN repos (similar to 13k revisions in audacity), I usually use git svn clone --stdlayout <svn_url> but it should be equivalent to your procedure
somewhere under .git you should see a file that changes during the fetch phase (IIRC it lists the grabbed revisions), so you will know it does something, also if svn:// url works, it should be faster than http://
Using the svn:// uri failed for me, so I'm back to using the http://. With --stdlayout will that capture the branches and tags information?
On 08/01/14 14:27, Darryl L. Pierce wrote:
The problem is that it's been HOURS and Git hasn't finished cloning the tree.
It's the third step which is taking forever to complete. I left it to run overnight (10 hours) and it never finished. I paused it, came to work, resumed it and it's still going.
Should this process really take this long?
of course it shouldn't, considering the state of the SCM art. but this is SVN, so that is entirely expected.
Or is there a faster, more efficient way to do this? I know that part of what'll make it take so long is the amount of history that has to be extracted from Subversion. But is that many hours normal or did I do something wrong?
perhaps you can prod upstream to migrate to something better.
btw if the clone is interrupted you can just continue with "git svn fetch", so at least you don't have to start from scratch.
oh, and consider yourself lucky that the only SVN repo _only_ has 13k commits and isn't just part of a single giant 100-project repo :)
On Wed, Jan 08, 2014 at 03:12:08PM +0100, Michael Stahl wrote:
On 08/01/14 14:27, Darryl L. Pierce wrote:
The problem is that it's been HOURS and Git hasn't finished cloning the tree.
It's the third step which is taking forever to complete. I left it to run overnight (10 hours) and it never finished. I paused it, came to work, resumed it and it's still going.
Should this process really take this long?
of course it shouldn't, considering the state of the SCM art. but this is SVN, so that is entirely expected.
Thanks for confirming, somewhat, that I'm not _totally_ insane. :D
Or is there a faster, more efficient way to do this? I know that part of what'll make it take so long is the amount of history that has to be extracted from Subversion. But is that many hours normal or did I do something wrong?
perhaps you can prod upstream to migrate to something better.
I may broach the subject if I become a regular committer with them. I can understand, though, why some projects stick with something like Subversion. My own work project is Subversion based, but we have some infrastructure in place to centrally maintain a git repo that updates every 15 minutes or so.
-
Dan Horák -
Darryl L. Pierce -
Michael Stahl