John McEleney / casperjs

With SlimerJS, checkStep turns into an infinite loop when
clicking on an <a target="_blank"> links, because navigationRequested
is staying always at true. In fact, it should not be modify
to true. In this case, onNavigationRequested is called with
isMainFrame=true instead of false. This is an issue
of SlimerJS, that needs a patch on Gecko.
https://github.com/laurentj/slimerjs/issues/48

This is a "bypass" patch, and should be removed when
SlimerJS will be released with its own patched Gecko.

Laurent Jouanneau authored 2013-06-24 16:37:15 +0200

When an assertion fails, casperjs stops any steps in the current suite.
This could cause issues in a test suite. For example, "withFrame"
generates some steps:

- one to switch to the given frame
- one to execute the given tests
- one to switch to the main frame.

If tests fail, the third step is not executed, and then next tests fail
because they are not executed in the main frame.

we can saw this issue in selftests: if a test fails in tests/suites/casper/frame.js,
tests in tests/suites/casper/global.js fail too.

Laurent Jouanneau authored 2013-06-24 12:24:42 +0200

Laurent Jouanneau authored 2013-06-24 11:32:12 +0200

- skip tests about javascript: links, because they are executed asynchronously
  so tests about them fails.
- event.x and event.y don't exist in Gecko. The standard way is event.clientX
  and event.clientY

Laurent Jouanneau authored 2013-06-21 10:37:51 +0200

Using instanceof is not enough for Gecko.

Laurent Jouanneau authored 2013-06-20 08:22:25 +0200

Laurent Jouanneau authored 2013-06-20 07:29:35 +0200

Laurent Jouanneau authored 2013-06-20 07:28:36 +0200

the navigationRequested boolean should be false when the new URL
has only a difference on the hash part with the previous URL, because
in this case we don't have a loadStarted/loadFinished event (at least
with Gecko) and then navigationRequested cannot be turn into false.
The consequence is that checkStep doesn't execute the next step because
of its test on navigationRequested.

Laurent Jouanneau authored 2013-06-20 07:21:51 +0200

In SlimerJS, the webpage object has no type 'qtruntimeobject' written
in the string form of the object. But it has a __type property.
The function should check this property when the engine is not PhantomJS.

Laurent Jouanneau authored 2013-06-18 20:24:34 +0200

It will easier in the code to know the current engine.

Laurent Jouanneau authored 2013-06-18 19:32:15 +0200

stackArray is not a standard property of the Error object. It is
added by PhantomJS (modification in the sources of QTWebkit).
We should use the stack property if stackArray is not present
(like in SlimerJS)

Laurent Jouanneau authored 2013-06-13 13:28:34 +0200

Laurent Jouanneau authored 2013-06-09 15:29:34 +0200

One test of clitests loads about:blank. However, there
is no navigationRequested event and so our new browserInitializing
flag is not toggled to false, and then checkStep() is blocking
tests. Let's toggle it inside onLoadStarted too.

Laurent Jouanneau authored 2013-06-09 15:13:58 +0200

PhantomJS has a strange behavior on fs.read: it seems it reads
file always as binary file. However SlimerJS's fs.read strictly
respect the 'b' flag. So the web server of selftest should
return content of binary files as... binary data. Else images
are not sent correctly to the browser and Gecko cannot display
them. and then some tests fail, like tests on assertVisible.

Laurent Jouanneau authored 2013-06-09 14:43:47 +0200

With Gecko, a warning appears when an option does not exists
in styles.

Laurent Jouanneau authored 2013-06-09 11:13:24 +0200

Laurent Jouanneau authored 2013-06-08 22:08:43 +0200

The call of webpage.open() (or openUrl()) is synchronous in PhantomJS,
but is asynchronous in SlimerJS (it needs to open a real window etc..).

So CasperJS should not launch the next step just after the call of
webpage.openUrl(). A flag is set to true before this call, and
then set to false when receiving the first event from the page loader.
And this flag has to be checked before to execute the next step.

Laurent Jouanneau authored 2013-06-08 00:10:54 +0200

Laurent Jouanneau authored 2013-06-05 00:43:53 +0200

Laurent Jouanneau authored 2013-06-02 22:48:25 +0200

Nicolas Perriault authored 2013-06-02 20:46:21 +0200

Nicolas Perriault authored 2013-06-02 20:31:09 +0200

Nicolas Perriault authored 2013-06-02 20:27:29 +0200

refs #490 - checking for ANSICON on colorizer.js to have colors on Windows

Nicolas Perriault authored 2013-05-28 03:41:21 -0700

renatodarrigo authored 2013-05-27 18:14:58 -0300

Check env variable 'ANSICON', and if exists, colorize the output.

renatodarrigo authored 2013-05-27 17:49:41 -0300

Check env variable 'ANSICON', and if exists, colorize the output.

renatodarrigo authored 2013-05-27 11:55:05 -0300

Nicolas Perriault authored 2013-05-27 09:21:28 +0200

spelling fix

Nicolas Perriault authored 2013-05-23 01:21:07 -0700

Tzvi Friedman authored 2013-05-22 22:32:49 -0300

Laurent Jouanneau authored 2013-05-23 01:49:33 +0200

This options allows to use SlimerJS (gecko) instead of PhantomJS (webkit)

Laurent Jouanneau authored 2013-05-21 22:23:19 +0200

It will allow to choice different browser engines, instead of phantomjs

Laurent Jouanneau authored 2013-05-21 15:03:00 +0200

SlimerJS provides an API to indicate paths where modules
can be found: require.paths (CommonJS specification).
So patchRequire is not needed for SlimerJS.

Laurent Jouanneau authored 2013-05-21 09:24:05 +0200

In a javascript context, all JS function are not all attached to the window object.
This is the case of 'native' functions. In SlimerJS, the 'require' function is
not attached to the window object. It is a standalone function.

Laurent Jouanneau authored 2013-05-21 09:23:45 +0200

Nicolas Perriault authored 2013-05-13 08:13:42 +0200

Nicolas Perriault authored 2013-05-13 08:02:42 +0200

Nicolas Perriault authored 2013-05-12 21:37:17 +0200

Nicolas Perriault authored 2013-05-10 14:54:27 +0200

Nicolas Perriault authored 2013-05-09 19:41:33 +0200

Nicolas Perriault authored 2013-05-09 19:03:43 +0200