Somebody wrote at http://wiki.archlinux.org/index.php/Tal … guidelines:
Any reason for using 'svn co' instead of 'svn export'? export is way faster for these situations and it doesn't create all the .svn files.
What do you think about that?
I agree...there's no reason that I can think of that we would need to do a full checkout rather than an export. Unless there is some sort of tool like versionpkg or whatever that might use that to its advantage when updating/rebuilding packages.
If you're using versionpkg, co is definatly faster than export. If you check out all the .svn stuff, subversion can just update the directory next time instead of checking out the entire thing again. I think this is why you'll usually see checked out code being moved to a build directory in svn PKGBUILDs, this allows svn to update all the source code, and still be able to delete the whole build directory to clean up after building.
I did some tests with checking out psi and virtualbox from svn. Here are results.
Connection with psi-im.org was not very stable, speed was changing between 64 and 128 kbytes/s, but average time (I did 5 tries: 3 with revision 600 and 2 with latest revision, I didn`t notice difference in time between 600 and 618 revisions) was nearly the same for both 'svn co' and 'svn export': ~1m30s.
When I was checking out virtualbox download speed always was 128 kbytes/s so I did only one try. Checkout time was absolutely the same - 4m21s.
psi svn co svn export download time ~1m30s hdd size (r600) 31516 kb 14520 kb hdd size (r618) 31576 kb 14540 kb download size ~8576 kb virtualbox svn co svn export download time 4m21s hdd size 212016 kb 94472 kb download size 33408 kb
Conclusion: there is no need in 'svn export'. If 'svn export' will be used somethere, those who want to checkout-and-delete will not get time boost, but those who want to keep checked out version will be forced to edit PKGBUILD`s.
Last edited by George_K (2007-04-22 13:46:02)