How are Jenkins builds killed exactly?

[Stephan Bergmann]

On 10/01/2020 15:57, Stephan Bergmann wrote:
> On 08/01/2020 16:23, Stephan Bergmann wrote:
>> I think I now understand what's going on:
> 
> ...and hope that <https://gerrit.libreoffice.org/c/lode/+/86522> "Make 
> sure tb_slave_wrapper kills all spawned processes" will get us towards a 
> fix

This is in place now for 
<https://ci.libreoffice.org/job/gerrit_linux_clang_dbgutil/> for a 
while, and appears to be effective so far.

Just now I also enabled it for 
<https://ci.libreoffice.org/job/lo_tb_master_linux_dbg/>, changing the 
build shell command at 
<https://ci.libreoffice.org/job/lo_tb_master_linux_dbg/configure> from

> ${LODE_HOME}/bin/tb_slave_wrapper --branch=MASTER --os=Linux --real --name=Jenkins_Linux_Dbg --distro-config="dev_master" --check --docs

to

> ${LODE_HOME}/bin/kill-wrapper "${LODE_HOME}/bin/tb_slave_wrapper --branch=MASTER --os=Linux --real --name=Jenkins_Linux_Dbg --distro-config=dev_master --check --docs"

(That job is restricted to tb75-lilith, so will have a recent enough 
lode repo with a known-working bin/kill-wrapper executable checked out.)

Build <https://ci.libreoffice.org/job/lo_tb_master_linux_dbg/28606/> had 
been aborted with

> Build timed out (after 45 minutes). Marking the build as aborted.

and the following 
<https://ci.libreoffice.org/job/lo_tb_master_linux_dbg/28607/> had 
failed UITest_chart in a way that smelled like a leftover soffice.bin 
again, and indeed there was a matching zombie on tb75 (which I killed 
manually).