We have a few plans that are CVS-based, set to build whenever the repository changes, with a polling frequency of 60. For whatever reason, these builds aren't locating changes anymore. After the import, the initial build-dir directories were empty, so I manually clicked 'Start Build' on each plan, and the code was checked out and the plan ran. Now it seems to not detect any more changes.

      A lot of things like this:

      2008-03-11 13:10:33,077 INFO [pool-13-thread-1] [InitialBuildListener] Changed detection request for LW-COMPILE ignored. Plan is currently busy (may already be building)

      .. are printed in the logs. No changes are found when CVS is polled. There are lingering lockfiles left on the CVS server that get re-added whenever Bamboo checks the changes, though, which prevents other people from updating/commiting.

            [BAM-2359] Plan Not Building

            I do not have the permission to see BSP-839. How and when will the solution be made available?

            Regards,

            Wim

            Deleted Account (Inactive) added a comment - I do not have the permission to see BSP-839. How and when will the solution be made available? Regards, Wim

            The issue that Ringo has reported is being handled as BSP-839.

            Adrian Hempel [Atlassian] added a comment - The issue that Ringo has reported is being handled as BSP-839 .

            Mark,

            I doubt if this issue is really fixed. I have a test setup with Bamboo 2.0.2 on a Windows 2000 machine, with CVS polling based build plans. Every rlog command doesn't return any results:

            {{22-mei-2008 09:09:11 Running CVS log command: "rlog -N -r -d2008-05-21 12:52:53 +0200< main "
            22-mei-2008 09:09:11 ... in: "C:/Data/bamboo/xml-data/build-dir/MAIN-HEAD/main"
            22-mei-2008 09:09:11 ... CVSROOT: ":pserver:username@ourcvshost:/dvlp/loc/cvs/repositories/tina"
            22-mei-2008 09:09:12 Found 0 change sets in the change logs}}

            If you look back in the history of this issue, Sam made a comment about the quotes and the trailing space in the cvs rlog command. I adapted the command with the info he made and got the following results now:

            {{$ cvs rlog -N -r -d"2008-05-20 12:00:00<" main
            cvs rlog: Logging main

            RCS file: /ourcvshost/dvlp/loc/cvs/repositories/tina/main/.cvsignore,v
            head: 1.2
            branch:
            locks: strict
            access list:
            keyword substitution: kv
            total revisions: 3; selected revisions: 0
            description:
            =============================================================================}}

            <cut much more log information>

            Can you look into this again?

            Ringo

            Deleted Account (Inactive) added a comment - Mark, I doubt if this issue is really fixed. I have a test setup with Bamboo 2.0.2 on a Windows 2000 machine, with CVS polling based build plans. Every rlog command doesn't return any results: {{22-mei-2008 09:09:11 Running CVS log command: "rlog -N -r -d2008-05-21 12:52:53 +0200< main " 22-mei-2008 09:09:11 ... in: "C:/Data/bamboo/xml-data/build-dir/MAIN-HEAD/main" 22-mei-2008 09:09:11 ... CVSROOT: ":pserver:username@ourcvshost:/dvlp/loc/cvs/repositories/tina" 22-mei-2008 09:09:12 Found 0 change sets in the change logs}} If you look back in the history of this issue, Sam made a comment about the quotes and the trailing space in the cvs rlog command. I adapted the command with the info he made and got the following results now: {{$ cvs rlog -N -r -d"2008-05-20 12:00:00<" main cvs rlog: Logging main RCS file: /ourcvshost/dvlp/loc/cvs/repositories/tina/main/.cvsignore,v head: 1.2 branch: locks: strict access list: keyword substitution: kv total revisions: 3; selected revisions: 0 description: =============================================================================}} <cut much more log information> Can you look into this again? Ringo

            Sam Berlin added a comment -

            Just want to confirm that 2.0b9 fixes the problem nicely.

            (You might want to keep in mind the issue w/ lingering lockfiles, incase the shift to log/rlog causes any other cvs users to run into problems. It was caused on our side by malformed ,v files in the Attic that led any log or rlog command that hit those files to fail.)

            Sam Berlin added a comment - Just want to confirm that 2.0b9 fixes the problem nicely. (You might want to keep in mind the issue w/ lingering lockfiles, incase the shift to log/rlog causes any other cvs users to run into problems. It was caused on our side by malformed ,v files in the Attic that led any log or rlog command that hit those files to fail.)

            MarkC added a comment -

            Sam,

            Thanks for the help on this!

            Beta 9 should be out tomorrow, please give it a try and see how you go

            Cheers,

            Mark C

            MarkC added a comment - Sam, Thanks for the help on this! Beta 9 should be out tomorrow, please give it a try and see how you go Cheers, Mark C

            I had a hunch that it could be returning quickly because there was no data to check on, so looked on the box in the build dir, this is what I see:

            -bash-3.00$ pwd; for A in `ls .`; do echo Looking in: $A; ls $A; done;
            /n/bin/bamboo/data_20b6/xml-data/build-dir
            Looking in: LW-COMPILE
            Looking in: LW-COMPILE15
            Looking in: LW-COMPONENTS
            Looking in: LW-MAGNET
            limewire
            Looking in: LW-MAVEN
            Looking in: LW-NIGHTLY
            limewire
            Looking in: LWS-DEF
            build-number.txt clover.license deploy documentation lib LICENSE.txt pom.xml scripts src target velocity.log

            It seems that after the import, no CVS dirs were created in plans that have a change-triggered CVS repository. LWS-DEF is svn-based, and LW-MAGNET & LW-NIGHTLY are scheduled. Interestingly, the first build of LW-NIGHTLY showed no changes, but the second did. (LW-MAGNET likely did the same thing, but I'm not sure.) LW-COMPILE, LW-COMPILE15, LW-COMPONENTS, and LW-MAVEN never found anything to build.

            The scheduled build found changes because it did a log on existing files the second night, whereas every other build (and scheduled on the first night) find no changes.

            Perhaps rlog should be used instead of log? log apparently requires that the files exist already, in order to check their log. I think this wouldn't work for new files or directories (or new files within new directories), as well as not working if nothing's checked out in the first place.

            Sam Berlin added a comment - I had a hunch that it could be returning quickly because there was no data to check on, so looked on the box in the build dir, this is what I see: -bash-3.00$ pwd; for A in `ls .`; do echo Looking in: $A; ls $A; done; /n/bin/bamboo/data_20b6/xml-data/build-dir Looking in: LW-COMPILE Looking in: LW-COMPILE15 Looking in: LW-COMPONENTS Looking in: LW-MAGNET limewire Looking in: LW-MAVEN Looking in: LW-NIGHTLY limewire Looking in: LWS-DEF build-number.txt clover.license deploy documentation lib LICENSE.txt pom.xml scripts src target velocity.log It seems that after the import, no CVS dirs were created in plans that have a change-triggered CVS repository. LWS-DEF is svn-based, and LW-MAGNET & LW-NIGHTLY are scheduled. Interestingly, the first build of LW-NIGHTLY showed no changes, but the second did. (LW-MAGNET likely did the same thing, but I'm not sure.) LW-COMPILE, LW-COMPILE15, LW-COMPONENTS, and LW-MAVEN never found anything to build. The scheduled build found changes because it did a log on existing files the second night, whereas every other build (and scheduled on the first night) find no changes. Perhaps rlog should be used instead of log? log apparently requires that the files exist already, in order to check their log. I think this wouldn't work for new files or directories (or new files within new directories), as well as not working if nothing's checked out in the first place.

            MarkC added a comment -

            I wonder if it's the SSH link that's causing the issue. We'll try to test it out and see how that goes.

            I'm running the pserver build locally and it seems to be picking up the changes (and building), albeit slowly.

            MarkC added a comment - I wonder if it's the SSH link that's causing the issue. We'll try to test it out and see how that goes. I'm running the pserver build locally and it seems to be picking up the changes (and building), albeit slowly.

            It's setup as..
            Repository: CVS
            CVS Root: :ext:bamboo@cvs.limewire.org:/cvs
            Authentication Type: SSH
            Private Key: /path/to/private_key

            The pserver repository is an rsync'd mirror that runs in a jail. I'll talk with the sysadmins about setting up a test user that can hit the SSH'd repo, if you need to.

            Sam Berlin added a comment - It's setup as.. Repository: CVS CVS Root: :ext:bamboo@cvs.limewire.org:/cvs Authentication Type: SSH Private Key: /path/to/private_key The pserver repository is an rsync'd mirror that runs in a jail. I'll talk with the sysadmins about setting up a test user that can hit the SSH'd repo, if you need to.

            MarkC added a comment -

            Sam,

            I'm trying to look into this further by replicating the LimeWire environment. Are you using the CVSROOT

            :pserver:guest@cvs.limewire.org:/cvs
            

            The strangeness you see (rogue spaces, no escaping) is only an artifact of the toString method in the CVS library we're using.

            One interesting thing in your logs is that the cvs log command returns very quickly.

            2008-03-23 09:38:33,410 INFO [pool-17-thread-1] [CvsRepositoryManager] Running CVS log command: "log -N -S -r -d2008-03-20 13:31:00 -0400< "
            2008-03-23 09:38:33,410 INFO [pool-17-thread-1] [CvsRepositoryManager]                  ... in:  "/n/bin/bamboo/data_20b6/xml-data/build-dir/LW-COMPILE15/limewire"
            ...
            2008-03-23 09:38:33,719 INFO [pool-17-thread-1] [CVSRepository] Found 0 change sets in the change logs
            

            That's a 300 millsecond return... In my test, the log commands take a significantly longer time to complete so I think there's something wrong here. I'm adding some optional extra logging that we can use.

            Cheers,

            Mark C

            MarkC added a comment - Sam, I'm trying to look into this further by replicating the LimeWire environment. Are you using the CVSROOT :pserver:guest@cvs.limewire.org:/cvs The strangeness you see (rogue spaces, no escaping) is only an artifact of the toString method in the CVS library we're using. One interesting thing in your logs is that the cvs log command returns very quickly. 2008-03-23 09:38:33,410 INFO [pool-17-thread-1] [CvsRepositoryManager] Running CVS log command: "log -N -S -r -d2008-03-20 13:31:00 -0400< " 2008-03-23 09:38:33,410 INFO [pool-17-thread-1] [CvsRepositoryManager] ... in: "/n/bin/bamboo/data_20b6/xml-data/build-dir/LW-COMPILE15/limewire" ... 2008-03-23 09:38:33,719 INFO [pool-17-thread-1] [CVSRepository] Found 0 change sets in the change logs That's a 300 millsecond return... In my test, the log commands take a significantly longer time to complete so I think there's something wrong here. I'm adding some optional extra logging that we can use. Cheers, Mark C

            Sam Berlin added a comment -

            I attached the logs to BSP-633

            Sam Berlin added a comment - I attached the logs to BSP-633

              mark@atlassian.com MarkC
              392fb6eb772b Sam Berlin
              Affected customers:
              0 This affects my team
              Watchers:
              2 Start watching this issue

                Created:
                Updated:
                Resolved: