I have a problem that is very likely not Pydev's fault, but I could use some insight from a developer. we use a clustered isilon server over NFS which does not support file locking (yeah, seems crazy to me to). this was not an issue with tiger, but with Leopard it is wreaking havoc. since the switch to Leopard eclipse takes about 5 minutes to open a workspace on our server, and the delay is directly related to pydev. The splashscreen says "Loading org.python.pydev" during most of that time. we launch eclipse with locking disabled (-Dosgi.locking=none) which was necessary to even open eclipse to a server workspace on leopard.
my test workspaces are completely empty of files and folders. i even blew away the .metadata folders to be safe.
- WITHOUT an interpreter setup in pydev:
opening workspace on local hard drive: fast
opening workspace on server: fast
- WITH an interpreter setup in pydev:
opening workspace on local hard drive: fast
opening workspace on server: 5 minutes
I was hoping that you might have some insight into what's going on on Pydev's side that might cause it to take so long in this situation. perhaps Pydev doesn't respect the -Dosgi.locking=none ?
thanks,
chad
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Pydev doesn't actually lock any files, so, this shouldn't be the problem... (I think that eclipse only uses the locking to make sure that only 1 instance of eclipse is using a given workspace at a time)
From your details it seems that it may be an issue restoring the information from the interpreter setup. I'm not sure the network speed you have there, but the configurations on my machine (when using pydev and pydev extensions) are around 16 MB and 4.000 files (with 4 interpreters configured), and that's all loaded in the startup (it could be lazier, but then again, the 1st time you do a completion operation or something else related, you'd have a halt, which could be perceived as worse), so, maybe just listing those contents and getting those contents could be responsible for that time?
Cheers,
Fabio
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
A good way to check that would be running the startup from a profiler and see where it's getting stuck...
The major reason for me would still be something on the network.
Aside from that, other things should be the same... maybe linux just deals better NFS than Mac OS -- so, looking for issues related to network slowness on Mac and NFS might be a good idea (you could try measuring the NFS communication speed with linux and with Mac OS to see if it's really slower from Mac OS than from linux).
Cheers,
Fabio
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I have a problem that is very likely not Pydev's fault, but I could use some insight from a developer. we use a clustered isilon server over NFS which does not support file locking (yeah, seems crazy to me to). this was not an issue with tiger, but with Leopard it is wreaking havoc. since the switch to Leopard eclipse takes about 5 minutes to open a workspace on our server, and the delay is directly related to pydev. The splashscreen says "Loading org.python.pydev" during most of that time. we launch eclipse with locking disabled (-Dosgi.locking=none) which was necessary to even open eclipse to a server workspace on leopard.
my test workspaces are completely empty of files and folders. i even blew away the .metadata folders to be safe.
- WITHOUT an interpreter setup in pydev:
opening workspace on local hard drive: fast
opening workspace on server: fast
- WITH an interpreter setup in pydev:
opening workspace on local hard drive: fast
opening workspace on server: 5 minutes
I was hoping that you might have some insight into what's going on on Pydev's side that might cause it to take so long in this situation. perhaps Pydev doesn't respect the -Dosgi.locking=none ?
thanks,
chad
Pydev doesn't actually lock any files, so, this shouldn't be the problem... (I think that eclipse only uses the locking to make sure that only 1 instance of eclipse is using a given workspace at a time)
From your details it seems that it may be an issue restoring the information from the interpreter setup. I'm not sure the network speed you have there, but the configurations on my machine (when using pydev and pydev extensions) are around 16 MB and 4.000 files (with 4 interpreters configured), and that's all loaded in the startup (it could be lazier, but then again, the 1st time you do a completion operation or something else related, you'd have a halt, which could be perceived as worse), so, maybe just listing those contents and getting those contents could be responsible for that time?
Cheers,
Fabio
Our network is very fast and the issue does not occur on linux, so it's something else. but i have no idea what.
A good way to check that would be running the startup from a profiler and see where it's getting stuck...
The major reason for me would still be something on the network.
Aside from that, other things should be the same... maybe linux just deals better NFS than Mac OS -- so, looking for issues related to network slowness on Mac and NFS might be a good idea (you could try measuring the NFS communication speed with linux and with Mac OS to see if it's really slower from Mac OS than from linux).
Cheers,
Fabio