Hudson β€œstops” / cancels work, leaving processes behind

We have an automatic assembly controlled by Hudson (1.312), which received several layers. On the slaves:

  • codewarrior cc / ld and other tools (some of them) executed
  • several recursive Cygwin GNUs are 3.81 instances executed
  • Hudson "execute batch windows command" (therefore cmd.exe runs the batch file created by hudson as close as possible)
  • A Hudson service running as a domain user (not a LocalService) started with
  • Windows XP (32-bit) machine

A wizard is an OSX machine running a Tomcat stock.

Problem: The cancel assembly button (the little red X that leads to the stop URL) does not return make or any of its subprocesses. As for the Hudson user interface, the job ended successfully, but the make.exe and cc / ld processes are still running. This can cause serious problems if the assembly starts again in the same workspace.

Is there a "best practice" to close all of these processes?

(Before anyone proposes a less Frankenstein build management structure such as ant, jam, maven, etc., we pretty successfully move on to jam! While we parse the excesses, we need killable assemblies on the current system.)

+3
source share
1

, , . ?

+5

Source: https://habr.com/ru/post/1712814/


All Articles