Can we either ensure the cell is scrolled into view before the value is read, or limit the number of columns we render in the affected tests (perhaps by splitting them into multiple tests)?
As a matter of fact, I'd rather do as you're suggesting. It seems like a more robust assertion and in general I'd like these feature specs to behave more reliably. Please wait for another patch to address this.
On Thu, May 3, 2018 at 11:05 AM Dave Page <dpage@pgadmin.org> wrote:
Hi
On Thu, May 3, 2018 at 3:23 PM, Anthony Emengo <aemengo@pivotal.io> wrote:
Hi Hackers,
The feature test at pgadmin4/web/pgadmin/feature_tests/view_data_dml_queries.py systematically fails in both our CI pipelines and when run locally. It seems to be because the selenium driver is unable to parse cells of the query result table when is it out of view. This issue seemed to have occurred after our most recent changes to how the feature tests run got committed into master.
Please consider the following patch to remedy this issue.
Thanks for looking at this - I was about to do so myself.
I'm not sure that maximising the window is the answer unfortunately. For example, the machine I'm hitting this with at the moment is a Mac which is essentially running headless (albeit connected to a KVM in a rack), which only wants to run at 800x600 resolution. I had hoped that the patch I committed yesterday to enable headless mode would solve that, but unfortunately not (it did solve some other issues I had running under a macOS Jenkins worker node, hence I committed it anyway). The other machine I hit this issue on is my travel laptop, which typically runs at 1024 x 640 or 1280 x 800 effective resolution.
Can we either ensure the cell is scrolled into view before the value is read, or limit the number of columns we render in the affected tests (perhaps by splitting them into multiple tests)?