scm command line questions and problems
1) If the workspace is open in Eclipse, I can't use scm:
C:\UserData\sedlock\rtc\workspace01\comp3>scm status
Could not initialize data area:
Another RCP application is running in this local workspace. File locked at file:C:\UserData\sedlock\rtc\workspace01
This is not good.
2) scm.exe seems to work fine in cygwin, except for the status command, which returns nothing:
************
sedlock@mucn13423 /cygdrive/c/UserData/sedlock/rtc/workspace01/comp3
$ scm.exe status
sedlock@mucn13423 /cygdrive/c/UserData/sedlock/rtc/workspace01/comp3
************
3) The doc says that if I login first, I should not be prompted for credentials, but I am always prompted for password.
C:\UserData\sedlock\rtc\workspace02>scm status
Password (sedlock @ https://localhost:9443/jazz/):
I can only use scm comfortably by supplying the password with -P.
4) There seems to be about a 20 second delay before anything happens after invoking scm. This is obviously a serious obstacle to using it in scripts where you may want to invoke it several times or in a loop.
I was going to benchmark RTC checkin vs. CC checkin with a loop that checks in files individually, which is perfectly fine with cleartool, when I realized that RTC was going to perform far worse than CC. I changed the script to check in the files in one go, and found that RTC checkin is, in my setup, about 3 times faster. But it is not always feasible to batch invocations of scm together.
I guess the overhead is the server login. I think you've got to implement some way of saving state so that invocations of scm within one login session do not require new server logins, but start to do their job immediately.
C:\UserData\sedlock\rtc\workspace01\comp3>scm status
Could not initialize data area:
Another RCP application is running in this local workspace. File locked at file:C:\UserData\sedlock\rtc\workspace01
This is not good.
2) scm.exe seems to work fine in cygwin, except for the status command, which returns nothing:
************
sedlock@mucn13423 /cygdrive/c/UserData/sedlock/rtc/workspace01/comp3
$ scm.exe status
sedlock@mucn13423 /cygdrive/c/UserData/sedlock/rtc/workspace01/comp3
************
3) The doc says that if I login first, I should not be prompted for credentials, but I am always prompted for password.
C:\UserData\sedlock\rtc\workspace02>scm status
Password (sedlock @ https://localhost:9443/jazz/):
I can only use scm comfortably by supplying the password with -P.
4) There seems to be about a 20 second delay before anything happens after invoking scm. This is obviously a serious obstacle to using it in scripts where you may want to invoke it several times or in a loop.
I was going to benchmark RTC checkin vs. CC checkin with a loop that checks in files individually, which is perfectly fine with cleartool, when I realized that RTC was going to perform far worse than CC. I changed the script to check in the files in one go, and found that RTC checkin is, in my setup, about 3 times faster. But it is not always feasible to batch invocations of scm together.
I guess the overhead is the server login. I think you've got to implement some way of saving state so that invocations of scm within one login session do not require new server logins, but start to do their job immediately.
11 answers
4) There seems to be about a 20 second delay before anything happens after invoking scm. This is obviously a serious obstacle to using it in scripts where you may want to invoke it several times or in a loop.
I was going to benchmark RTC checkin vs. CC checkin with a loop that checks in files individually, which is perfectly fine with cleartool, when I realized that RTC was going to perform far worse than CC. I changed the script to check in the files in one go, and found that RTC checkin is, in my setup, about 3 times faster. But it is not always feasible to batch invocations of scm together.
I guess the overhead is the server login. I think you've got to implement some way of saving state so that invocations of scm within one login session do not require new server logins, but start to do their job immediately.
David,
We are also piloting RTC, and had the same issue about the scm startup performance. See the messages near the end of thread -- http://jazz.net/forums/viewtopic.php?t=5987 The summary is that they are looking @ improving the startup time in the next couple of versions.
Regards,
Mike Johnson
This is good to know, but the posting you refer to seems to think the overhead is all due to java. I speculate there is also the overhead of a new login to the server. It would be good to get a confirmation about this. In any case a speedup from 20 to 14 seconds is not yet in the ballpark...
I don't presume to think for the developers, but to me it's fairly clear that scm has to be implemented in a low-overhead language (e.g. C) and it should simply send the command over to a client proxy, which is started with "scm login" and remains logged in to the RTC server till "scm logout". It in turn sends the command over to the server and funnels the output back to scm. I can't see any other way to avoid both Java and login overhead.
BTW, if there is an RTC Eclipse instance running on the client, maybe it could act as the preferred proxy, thus completely avoiding overhead.
David,
We are also piloting RTC, and had the same issue about the scm startup performance. See the messages near the end of thread -- http://jazz.net/forums/viewtopic.php?t=5987 The summary is that they are looking @ improving the startup time in the next couple of versions.
Regards,
Mike Johnson
I don't presume to think for the developers, but to me it's fairly clear that scm has to be implemented in a low-overhead language (e.g. C) and it should simply send the command over to a client proxy, which is started with "scm login" and remains logged in to the RTC server till "scm logout". It in turn sends the command over to the server and funnels the output back to scm. I can't see any other way to avoid both Java and login overhead.
BTW, if there is an RTC Eclipse instance running on the client, maybe it could act as the preferred proxy, thus completely avoiding overhead.
4) There seems to be about a 20 second delay before anything happens after invoking scm. This is obviously a serious obstacle to using it in scripts where you may want to invoke it several times or in a loop.
I was going to benchmark RTC checkin vs. CC checkin with a loop that checks in files individually, which is perfectly fine with cleartool, when I realized that RTC was going to perform far worse than CC. I changed the script to check in the files in one go, and found that RTC checkin is, in my setup, about 3 times faster. But it is not always feasible to batch invocations of scm together.
I guess the overhead is the server login. I think you've got to implement some way of saving state so that invocations of scm within one login session do not require new server logins, but start to do their job immediately.
David,
We are also piloting RTC, and had the same issue about the scm startup performance. See the messages near the end of thread -- http://jazz.net/forums/viewtopic.php?t=5987 The summary is that they are looking @ improving the startup time in the next couple of versions.
Regards,
Mike Johnson
1) If the workspace is open in Eclipse, I can't use scm:
That is correct. That will likely change further down the line - possibly in 2.0.0.2 or 3.0. I've submitted bug 90047 to track this one.
2) scm.exe seems to work fine in cygwin, except for the status command, which returns nothing:
Please open an RTC bug and file it against Source Control/Client Library/CLI.
3) The doc says that if I login first, I should not be prompted for credentials, but I am always prompted for password.
'scm login' doesn't save your password unless you pass the -P option in. That's changing with 2.0.0.1 - you'll be able to supply a '-c' which will tell the CLI to save your password no matter what.
I've submitted doc bug 90044.
4) There seems to be about a 20 second delay before anything happens after invoking scm. This is obviously a serious obstacle to using it in scripts where you may want to invoke it several times or in a loop.
There are some experimental features going in 2.0.0.1 that should improve start and run times of the CLI. In the meantime, (as mentioned elsewhere) try running with other VMs. You should see an incremental performance improvement.
e
'scm login' doesn't save your password unless you pass the -P option in. That's changing with 2.0.0.1 - you'll be able to supply a '-c' which will tell the CLI to save your password no matter what.
But I do use the -P argument with login. Here is the cmd I use:
scm.exe login -r https://mucn13423.eu.infineon.com:9443/jazz -u sedlock -P sedlock -n rtc
That looks correct. Then afterwards anywhere you see a command that
requires a -r, you should be able to pass in "rtc" because that is the
nickname you have specified, and it will use the uri, user name and
password you have specified.
Please file a bug if you're finding this is not the case.
On Fri, 14 Aug 2009 04:23:05 -0400, David.Sedlock.infineon.com
<David> wrote:
--
requires a -r, you should be able to pass in "rtc" because that is the
nickname you have specified, and it will use the uri, user name and
password you have specified.
Please file a bug if you're finding this is not the case.
On Fri, 14 Aug 2009 04:23:05 -0400, David.Sedlock.infineon.com
<David> wrote:
'scm login' doesn't save your password unless you pass the -P
option in. That's changing with 2.0.0.1 - you'll be able to supply a
'-c' which will tell the CLI to save your password no matter what.
But I do use the -P argument with login. Here is the cmd I use:
scm.exe login -r https://mucn13423.eu.infineon.com:9443/jazz -u
sedlock -P sedlock -n rtc
--
But I do use the -P argument with login. Here is the cmd I use:
scm.exe login -r https://mucn13423.eu.infineon.com:9443/jazz -u sedlock -P sedlock -n rtc
That's a bug. On initial login, could you verify that a line is added to ~/.jazz-scm/repositories.txt?
If everything is working, it should look like:
https://mucn13423.eu.infineon.com:9443/jazz,rtc,sedlock,garble
If your password isn't being stored, it will look like:
https://mucn13423.eu.infineon.com:9443/jazz,rtc,sedlock,
If your password isn't being stored, please open a bug and assign it to me.
e
If everything is working, it should look like:https://mucn13423.eu.infineon.com:9443/jazz,rtc,sedlock,garble
It looks exactly like this except for "garble".
Please run 'scm logout -r rtc', then log in again with the full set of credentials (ie, include the -P as you alluded to above). Do you get the full record (ie, with garble)?
If not, please submit a bug. I'd love to see this reproduced and reproducible. I really can't see what would be causing it.
e
Was this issue resolved in 2.0.0.2? This discussion sites the WI below, which says it was resolved in 2.0.0.2
https://jazz.net/jazz/web/projects/Rational%20Team%20Concert#action=com.ibm.team.workitem.viewWorkItem&id=90047
However, if I 'scm status' a repo workspace loaded using the Eclipse client, I see the original poster's error. If I halt Eclipse, the command works fine. In 2.0.0.2, can we mix Eclipse and command line scm?
D:\workspaces\workspace\Docs>scm status
Could not initialize data area:
Another RCP application is running in this local workspace. File locked at file: D:\workspaces\workspace
https://jazz.net/jazz/web/projects/Rational%20Team%20Concert#action=com.ibm.team.workitem.viewWorkItem&id=90047
However, if I 'scm status' a repo workspace loaded using the Eclipse client, I see the original poster's error. If I halt Eclipse, the command works fine. In 2.0.0.2, can we mix Eclipse and command line scm?
D:\workspaces\workspace\Docs>scm status
Could not initialize data area:
Another RCP application is running in this local workspace. File locked at file: D:\workspaces\workspace
page 1of 1 pagesof 2 pages