Currently we have two 64-bit bots, one building debug and the other building release. Apart from that they both run all the tests. We need to have separate bots for building and testing in order to detect build and test breakages easily. With the current setup is quite difficult to know the status of the port.
Created attachment 231005 [details] Patch
Patch attached
Comment on attachment 231005 [details] Patch Thanks for the patch!. Why are we bringing back the WK2 thing? There's no longer WebKit1 API in GTK+ port, so I think we can keep the names of the bots without the WK2 part.
Comment on attachment 231005 [details] Patch I think with the split we are missing the GObject DOM bindings API breaks step. We need to add the step RunGtkWebKitGObjectDOMBindingsAPIBreakTests to BuildAndTestFactory when port is gtk
(In reply to comment #4) > (From update of attachment 231005 [details]) > I think with the split we are missing the GObject DOM bindings API breaks step. We need to add the step RunGtkWebKitGObjectDOMBindingsAPIBreakTests to BuildAndTestFactory when port is gtk I meant to TestFactory
Created attachment 231060 [details] Patch
Thanks for the comments. I have addressed them and uploaded a new version of the patch.
Comment on attachment 231060 [details] Patch ok, let's try. Thanks!
Comment on attachment 231060 [details] Patch Clearing flags on attachment: 231060 Committed r168471: <http://trac.webkit.org/changeset/168471>
All reviewed patches have been landed. Closing bug.
This broke the master. Bug https://bugs.webkit.org/show_bug.cgi?id=132688 follows the issue
There was another problem with the original path (the builder names were not updated correctly on the scheduler list). Tracked here: https://bugs.webkit.org/show_bug.cgi?id=132699
(In reply to comment #0) > Currently we have two 64-bit bots, one building debug and the other building release. Apart from that they both run all the tests. We need to have separate bots for building and testing in order to detect build and test breakages easily. With the current setup is quite difficult to know the status of the port. Are you sure if it is efficient to split the builder and the tester for the debug bot? Uploading gigabytes to the master by builder and then download by the tester will take long time, maybe an hour per build if your bots aren't very close to build.webkit.org.
(In reply to comment #13) > > Are you sure if it is efficient to split the builder and the tester for the > debug bot? Uploading gigabytes to the master by builder and then download > by the tester will take long time, maybe an hour per build if your bots > aren't very close to build.webkit.org. We are planning to workaround this issue by wrapping the zip/unzip commands with a shell script that will copy the built product to the test buildslave directly (both machines are on the same LAN so it should be pretty fast). And instead of uploading the built product, the script will upload to the buildmaster a fake zip file that contains only a txt with a link to the real build product. This way we will avoid the problem with the slow upload/download. Probably it will take some few days of try-error-fix until we fully stabilize this new setup. Once the deployment is working as expected I plan to send a mail to the ML webkit-gtk@lists.webkit.org announcing this.
(In reply to comment #14) > Once the deployment is working as expected I plan to send a mail to the ML webkit-gtk@lists.webkit.org announcing this. BTW, https://lists.webkit.org/pipermail/webkit-gtk/2014-May/001909.html