Some svn numbers

[Daniel Berlin <dberlin at dberlin dot org>]

So, Mike Stump said his update times with svn from toolchain.org were a
little slow.

I posited a few reasons why this might be, and just to publicly follow
up with some numbers, here is what i get from gcc.gnu.org after
remembering to convert with keyword support turned off :)

These are all do-nothing updates of the improved-aliasing branch, from
gcc.gnu.org, same remote partition (IE they are both on the same disk on
gcc.gnu.org).  Real updates are faster with svn than cvs in  all cases i
tried.

Checkouts will be about 30% slower with svn, just because it has to
write more data out to disk because of the working copy (those who want
no/faster working copy are free to use svk, or wait for 1.4. 1.3 was
scheduled to allow compressed/no working copy checkouts, but the Summer
of Code student didn't end up doing it :( ).

(I changed my hostname recently, and haven't restarted, hence the xauth
crap)

dberlin@linux:/mnt/gccstuff/cvs/gcc> time cvs -q update -dP
Warning: No xauth data; using fake authentication data for X11
forwarding.

real    0m38.089s
user    0m2.420s
sys     0m3.056s

dberlin@linux:/mnt/gccstuff/cvs/gcc> time cvs -q update -dP
Warning: No xauth data; using fake authentication data for X11
forwarding.

real    0m52.082s
user    0m2.392s
sys     0m3.328s

dberlin@linux:/mnt/gccstuff/cvs/gcc> time cvs -q update -dP
Warning: No xauth data; using fake authentication data for X11
forwarding.

real    0m38.181s
user    0m2.420s
sys     0m2.828s
dberlin@linux:/mnt/gccstuff/cvs/gcc>
                                             
(I don't know what to explain the middle one with either, but hey, such
is life :P)

dberlin@linux:/mnt/gccstuff/svn/improved-aliasing-branch> time svn
update
Warning: No xauth data; using fake authentication data for X11
forwarding.
At revision 83795.

real    0m57.799s
user    0m2.872s
sys     0m1.712s
dberlin@linux:/mnt/gccstuff/svn/improved-aliasing-branch> time svn
update
Warning: No xauth data; using fake authentication data for X11
forwarding.
At revision 83795.

real    0m19.425s
user    0m1.712s
sys     0m0.644s
dberlin@linux:/mnt/gccstuff/svn/improved-aliasing-branch> time svn
update
Warning: No xauth data; using fake authentication data for X11
forwarding.
At revision 83795.

real    0m19.687s
user    0m1.992s
sys     0m0.724s
dberlin@linux:/mnt/gccstuff/svn/improved-aliasing-branch> cd ../../cvs/

So the first svn update took 57 seconds, and did nothing
The second two took 19 seconds each.

I posit this is because:
1. the first time svn ran, it had to stat a lot of files, and this
wasn't in the stat cache.
It then asked went out for the newest revision number of the repo, the
repo said "83795", and it said "Hey, cool, i'm done"
2. The next two times, the stat'ing took less time becuase it was in
cache, and the network time was negligible.


CVS was probably getting eaten by network traffic one of the times i ran
it, plus possibly stat times.
Who knows. :)


Time to switch to HEAD from my improved-aliasing-branch 
time cvs -q update -rHEAD -dP
...
real    2m28.183s
user    0m55.679s
sys     0m25.410s

time svn switch svn+ssh://gcc.gnu.org/export/u0/gccsvn/gccrepo/trunk

real    2m34.380s
user    0m12.429s
sys     0m13.885s




Mike, if you have shell access to gcc.gnu.org, try svn co svn
+ssh://gcc.gnu.org/export/u0/gccsvn/gccrepo/trunk

Let me know if it's still slow for you.

(I'd recommend a 1.2.x client, or 1.3.x in a few weeks.  1.2.x has much
fsater working copy than 1.2.x)