History | Log In     View a printable version of the current page.  
Issue Details (XML | Word | Printable)

Key: CIB-1930
Type: Bug Bug
Status: Resolved Resolved
Resolution: Fixed
Priority: 3 3
Assignee: jason
Reporter: Rohan McGovern
Votes: 0
Watchers: 0
Operations

If you were logged in you would be able to see more operations.
Pulse

Cancelling build during git bootstrap phase leads to errors in subsequent bootstraps

Created: 08/Apr/09 11:07 PM   Updated: 16/Apr/09 08:49 PM
Component/s: None
Affects Version/s: 2.0.26
Fix Version/s: 2.1.4, 2.0.28

Original Estimate: Unknown Remaining Estimate: Unknown Time Spent: Unknown


 Description  « Hide
We have observed that if we cancel a build while it is doing a git bootstrap, git processes are sometimes left running on the agent, and these prevent further bootstraps from working (most usual symptom: hanging indefinitely in the next bootstrap).

We have observed this on at least Windows Server, Windows XP, and SuSE Linux 11.1 x86_64.

A workaround is to manually log into the affected machines and kill any left over git processes.

 All   Comments   Work Log   Change History      Sort Order:
jason - 09/Apr/09 01:59 PM
Hi Rohan,

Thanks for this report. Process killing using the Java APIs is pretty primitive, so if git creates a process tree when it runs then some of the tree may be left around. We have observed this mostly on Windows, however, so a case on SuSE is different. We will likely need to write native code to clean up extra processes to fix this.

Rohan McGovern - 14/Apr/09 07:41 PM
I may have been premature in saying that we've observed it on SuSE Linux 11.1 x86_64. When this issue was last observed, I logged into the Windows machines and verified that there were child processes left over. Since the Linux machine was failing in the same way I assumed that was the same case, but did not actually ssh into the machine and verify.

Since then we have been seeing some unrelated hangs in bootstrap regardless of whether or not a build was canceled, so the hang on SuSE at that time may have been a coincidence.

jason - 16/Apr/09 10:08 AM
Hi Rohan,

Thanks for the update. In this case we may already have a solution which is just not turned on by default -- native code to improve process killing on Windows. We have decided to enable this code by default in the next build which will hopefully solve the issue. If you find it does not, then please reopen the issue.

jason - 16/Apr/09 08:49 PM
Merged to trunk in change 5837.