-----Original Message-----
From: David Kramer [mailto:david@thekramers.net]
- I don't want to do an svn copy every day for one of these builds,
because I think that would make the repository huge. I know it's a
"cheap" copy, but cheap != free, and 50 or 60 cheap copies of thousands of
files for every release can add up.
--------------------------
Well, I'd expect the cost of a copy to be several tens of bytes.
Multiplying 20 bytes/copy * 60 copies/file * 10,000 files = 12 megabytes,
which doesn't seem to be much these days. (And most of the files do not
need to be mentioned individually, as other posters have noted.)
--------------------------
I expect it would slow down access after a while, too.
--------------------------
I would hope not. Well-designed data structures usually have access time
proportional to at most the log of the number of items in the data
structure, so you might be adding 50% to the file-lookup time, not 50 times.
--------------------------
- I played with the idea of having a special file in the repository that
gets revision/label property pairs assigned to it, but apparently you
cannot get the properties of a file by URL, only working copy.
--------------------------
"svn propget svn:externals https://scm.sipfoundry.org/rep/sipXpbx/main"
worked for me.
-----------------
-I wrote a script that is beautiful in [its] hackishness. The heart of
getrevlabel.sh is:
svn cat $SVNREPOS/build/revlabels.txt | \
awk -v LABEL="$1" '$1==LABEL {print $2}'
Now the user can svn checkout -r `getrevlabel.sh 8.0a` $SVNREPOS
-----------------
function svn-checkout-label()
{
svn checkout -r $( svn cat $SVNREPOS/build/revlabels.txt | \
awk '$1=="'$1"' {print $2}' $) $SVNREPOS
}
The user can do "svn-checkout-label 8.0a".
Dale
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Dec 30 17:58:15 2004