On 24 Feb 2016, at 11:21, Sven Van Caekenberghe <sven@stfx.eu> wrote:
On 24 Feb 2016, at 11:09, Max Leske <maxleske@gmail.com> wrote:
Hereâs an idea:
1. exclude Zinc tests from the validation tests 2. after the build, trigger a Travis build on Github via API (I just set that up for Fuel, so I can provide help there) 3. the Travis build only runs the Zinc tests 4. read build results from Travis
Very ugly, I know. But itâs done rather quickly and should solve all the network problems.
I don't think the current problem is severe enough to put much work in it, right now.
But would it not be much better to run everything (all tests) on Travis ?
At first, maybe just as a limited experiment ? I would love to see that.
Sure. That wouldnât be hard (although I think support for Windows is missing for the Smalltalk language, but Iâm sure Fabio Niephaus would help us out there). We could use the Jenkins job as a trigger (or create a second job for experimenting first). In the long run, we probably would want to use the push / pull request hook to trigger the build but for quickly hacking things together I suggest using the API trigger. I donât have enough permissions on Github and Jenkins to pull this off alone but Iâd be happy to help setting up the Travis stuff and providing the trigger script.
Max
On 24 Feb 2016, at 11:01, Marcus Denker <marcus.denker@inria.fr> wrote:
The problem is that managing a CI server for a project like Pharo would be one full time engineer in a company, we do not have the manpower.
So we need to find solutions that are cheap to do.
On 24 Feb 2016, at 10:50, Ben Coman <btc@openinworld.com> wrote:
Maybe there can be a pre-test run at the shell level to flag that the required network connectivity is available to run that test inside the image. Pharo startup could read them in while starting. cheers -ben
On Wed, Feb 24, 2016 at 5:38 PM, Marcus Denker <marcus.denker@inria.fr> wrote:
one idea could be to add this to the filter of the CI runner.
It seems it fails due to network setup problems that are specific to the CI server...
On 24 Feb 2016, at 09:07, Sven Van Caekenberghe <sven@stfx.eu> wrote:
The following test seems to be failing a lot lately on the CI infrastructure, yet it always succeeds for me on my machine. Is there anybody who sees this fail on their machines ?
On 24 Feb 2016, at 08:36, no-reply@ci.inria.fr wrote:
https://ci.inria.fr/pharo/job/Pharo-5.0-Update-Step-2.1-Validation-M-Z/label...
1 regressions found. Zinc.Zodiac.ZnHTTPSTests.testAmazonAWS