Re: Intermittent Builder Windows Petri Dish x64 failed (failed ctest) failures seriously annoying.

[Gerald Combs <gerald () wireshark org>]

The failures don't appear to be limited to the test steps. Earlier we had the following cluster of failures:

x86 builder:

Branch        Build     Step           Start        Stop
master         1680     ctest       08:38:18    09:02:45
master-3.2      365     ctest       08:27:02    08:55:58
master-3.0      133     WiX pkg     07:56:04    08:16:38
master-2.6      221     NSIS pkg    07:43:59    08:05:01

x64 builder:

Branch        Build     Step           Start        Stop
master         1604     ctest       08:22:58    08:47:45
master-3.2      369     ctest       08:24:09    08:48:15
master-3.0      322     NSIS pkg    07:36:14    07:57:08
master-2.6      212     WiX pkg     07:35:23    07:59:54

Each builder is an EC2 instance at AWS, and the AWS console shows that the CPU on each instance was 100% during that 
time. The obvious fix is more and/or faster Windows builders), but I also noticed that tshark does more work than it 
should in many cases. If I open Process Monitor and set the filters

- Operation contains "Process Create"
- Path contains "<my Wireshark development folder>"

and run test/test.py, procmon shows that we create 5573 process for 576 tests. For the decryption suite,  procmon shows 
970 processes for 89 tests. Most of the created decryption processes are extcaps, which seems odd for those particular 
tests. If I rename the extcap directory, procmon shows that 97 process are created and the suite runs much more quickly 
(50s vs 93s).

On 7/2/20 8:02 AM, Graham Bloice wrote:
> The tests are run all in parallel, and I'm not sure if the PD is a spot or reserved instance which might affect 
> available CPU cycles.
>
> On Thu, 2 Jul 2020 at 15:58, Dario Lombardo <lomato () gmail com <mailto:lomato () gmail com>> wrote:
>
>     You're right, it's not the test I pointed out. However it worked fine before, then I would check the host to see 
> if it's suffering.
>
>     On Thu, Jul 2, 2020 at 4:51 PM Graham Bloice <graham.bloice () trihedral com <mailto:graham.bloice () trihedral 
> com>> wrote:
>
>         The decryption test might need a bigger timeout, currently at 600 seconds:
>
>         9/34 Test #3: suite_decryption .......................***Timeout 630.47 sec
>
>
>         On Thu, 2 Jul 2020 at 15:39, Anders Broman via Wireshark-dev <wireshark-dev () wireshark org 
> <mailto:wireshark-dev () wireshark org>> wrote:
>
>             Hi,____
>
>             We have intermittent false petri dish failures. Can someone with insight into the tests have a look?____
>
>             Best regards____
>
>             Anders____
>
>             Example Builder Windows Petri Dish x64 failed (failed ctest) - 
> http://buildbot.wireshark.org/petri-dish/builders/Windows%20Petri%20Dish%20x64/builds/11913____
>
>
>
>     -- 
>
>     Naima is online.
>
>
>
>
> -- 
> Graham Bloice
>
> ___________________________________________________________________________
> Sent via:    Wireshark-dev mailing list <wireshark-dev () wireshark org>
> Archives:    https://www.wireshark.org/lists/wireshark-dev
> Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev
>              mailto:wireshark-dev-request () wireshark org?subject=unsubscribe

___________________________________________________________________________
Sent via:    Wireshark-dev mailing list <wireshark-dev () wireshark org>
Archives:    https://www.wireshark.org/lists/wireshark-dev
Unsubscribe: https://www.wireshark.org/mailman/options/wireshark-dev
             mailto:wireshark-dev-request () wireshark org?subject=unsubscribe