Just an update for anyone who finds themselves having the same kind of
problems as me in future.
I moved all my files locally, and installed an eclipse plugin called
FileSync to automatically transfer changed files to the network share.
It works like a dream! As promised, Eclipse refreshes happen in a couple
of seconds. As a result, Subclipse switches, updates, and merges all
happen nearly as fast.
So the problem comes down to Eclipse being extremely slow to refresh a
project over a network share. I'm not certain why this would be. I have
a 1gbit ethernet connection to the fileserver. And the fileserver has
plenty of grunt (8 cores or something, and a HDD to match). I suspect it
is down to Samba and having to check file modified times/CRCs on so many
small files. There must be some kind of system hooks that Eclipse can
use to make this extremely fast from local disk that it can't use over
samba, is my guess. Any eclipse devs on the list care to weigh in?
Anyway, it probably will not make much difference for subclipse to not
do the full project refresh.
Thanks for all you help and advice Mark. And thanks for making a great
plugin! I'm certainly very happy now I can take full advantage of it.
Tom Walter wrote:
>> I have to think that Eclipse is reasonably efficient at doing a
>> refresh. You can have a large project on local disk and if you
>> refresh it and it is already in synch it runs in just a second or two.
>> So I do not think it can be doing a full crawl of all of the files
>> (although I do not see how else it could do it).
> A second or two! Wow if it should be running that fast there must be
> something seriously wrong with my setup. Perhaps Eclipse itself has a
> problem working on network drives? Or perhaps its a windows thing...
> eclipse might be able to hook into some OS-level thing when working
> locally that it can't use when accessing a network drive.
> Its so hard to tell with Eclipse, since it could be any one of a number
> of plugins that are causing the refresh to be slow, and I have ended up
> complaining about subclipse simply because that is the most visibly
> affected function for me.
>> So the question is will it be significantly faster for us to crawl the
>> structure and gather all the .svn folders and then refresh each of
>> those, versus just letting Eclipse do it all itself? Given that in
>> the switch case, more than half the total amount of files still need
>> to be processed, it seems likely to not be much faster to me. I think
>> it is worth trying if someone wants to setup the tests. The patch is
>> fairly easy.
> Is it just a matter of checking out the subclipse source, applying the
> patch. Running some java compile command, and dropping a JAR somewhere?
> If it's that simple I can give it a shot, as long as you can tell me
> what the compile command should look like and where to put the JAR.
> To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subclipse.tigris.org].
To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subclipse.tigris.org].
Received on 2009-01-16 01:45:10 CET