RFC: Pull request adding Travis CI support to Sproutcore

[Umberto Nicoletti]

Dear all,

this weekend I have been happily hacking away at a Travis CI integration for Sproutcore.

The result is this pull request:

https://github.com/sproutcore/sproutcore/pull/1025

Let me know what you think.

Ciao,

Umberto

[Umberto Nicoletti]

This is an update after the discussion on the last IRC meeting.

Details of this pr (what it does and how):

https://github.com/sproutcore/sproutcore/pull/1025

As you might know, the travis-ci integration requires this pull request:

https://github.com/sproutcore/sproutcore/pull/1018

which introduces support in Sproutcore for detecting phantomjs, the headless browser used by the countinuos integration

scripts. Phantomjs is actually Webkit, but compiled so that it can run headless, or without a graphical user interface.

The, imho obvious, advantage here is that *all* unit tests are run on each push and pull request, thus greatly simplifying the work

of reviewers, coders, and also making sure there are no side effects with each change.

Someone (rightly) objects at having and maintaining a specific detector for Phantomjs in Sproutcore, but I would say the the benefits

greatly outweigh the costs of maintaining a single line of code and there is really no other readily available way of working around this issue.

Anyway, to prove that #1018 does no harm I have run all unit tests with and without it

and I have found out that that change singlehandedly fixes 7 test suites with no regressions.

The results are available on this Google spreadsheet:

https://docs.google.com/spreadsheet/pub?key=0Ak5_P3HeN3GEdHBMaXRFNTF5TFRVbkExQjMyVmxlRVE&output=html

The FIXED suites are those fixed by #1018, those FAILED are, well, still failing for other reasons, and they also fail when run in my browser. I have quickly dug

through the FAILED ones and fixed a couple of them. The commits are available in my fork and are linked to in the last spreadsheet column.

Summary: the patch fixes 7 test suites with no regressions. There are still 27 test suites with 190 assertion failures and 18 errors.

Of the 27 suites 2 are related to templateviews and two I have resolved myself.

Why should we merge this before 1.10?

Because automated CI saves time and gives immediate feedback, when it's most effective.

Because Tyler Keating's huge work on the refactoring views and introducing statecharts for views is large, deep reaching into SC internals and therefore deserves an automation to guide him and to reduce his workload.

Ciao,

Umberto

[Dave Porter]

I merged the window._phantom check into SC.platform this morning. Given that the Travis-CL integration doesn't appear to require any other changes to SproutCore itself, I agree that we should merge this in ASAP.

You mentioned that you're fixing some of the other failing tests, is that work available anywhere?

--
You received this message because you are subscribed to the Google Groups "SproutCore Developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sproutcore-de...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Nicolas BADIA]

Just wanna say BRAVO Umberto !

This is a really cool addition.

----

Nicolas

[Umberto Nicoletti]

Il giorno lunedì 5 agosto 2013 18:38:55 UTC+2, Dave ha scritto:

I merged the window._phantom check into SC.platform this morning. Given that the Travis-CL integration doesn't appear to require any other changes to SproutCore itself, I agree that we should merge this in ASAP.

You mentioned that you're fixing some of the other failing tests, is that work available anywhere?

it's referenced in this google doc:

https://docs.google.com/spreadsheet/pub?key=0Ak5_P3HeN3GEdHBMaXRFNTF5TFRVbkExQjMyVmxlRVE&output=html

look at the last column on the right and then scroll down, you should find three links to commits on my fork.

I planned to open 3 pull requests as soon as this was merged, so I should be able to push them shortly.

BR,

Umberto