|
»
»
« Return to Thread: [1.02] ReferenceError: "navigator" running tests in Rhino
Re: [1.02] ReferenceError: "navigator" running tests in Rhino
by James Burke-3
::
Rate this Message:
On Dec 27, 2007 9:34 AM, Kalle Korhonen <kalle.o.korhonen@...> wrote:
> On 12/24/07, James Burke <jburke@...> wrote:
>
> > The built version that you get from the 1.0.2 download location (not
> > the -src file) is built for the browser, and not for rhino. Running
> > that version with rhino and DOH can produce the "navigator" issue.
>
>
> Thanks a lot James for your answers, this is really helpful. Quickly looking
> at the build.xml, I didn't see a target that would specifically "build for
> the browser" and I couldn't find any documentation about it. What target
> should I run to create this version? And why is it needed - wasn't one of
> the design goals of dojo 1.0 to target purely a browser environment?
The default behavior of the build system is to build for the browser.
Looking more at it now, the build system has code in it to build a
rhino-specific build, but that build process fails right now. So usage
in rhino right now means just using a source distribution. I filed
this trac bug for doing a build for the rhino environment (I just
fixed it in the repository trunk):
http://trac.dojotoolkit.org/ticket/5467
See that ticket for an example profile file that makes a build for the
rhino environment (requires the latest trunk code to work).
> > Dojo can be run in the browser or in rhino, but the built versions we
> > put on the website are only for running in the browser. We do not
> > support auto-detection of the environment in the built version to cut
> > down on the size of dojo.js and the vast majority of users only care
> > about the browser version. So, for running in rhino, you will need to
> > get the -src download or grab the source from subversion.
>
> It'd be good to document this at
> http://dojotoolkit.org/book/dojo-book-0-9/part-4-meta-dojo/d-o-h-unit-testing
> (or presumably upcoming dojo-book-1-0).
I put a note near the Rhino testing section about using a source
distribution when running DOH in Rhino.
> > DOH does not provide a simulation of the browser environment in rhino.
> > Instead, it runs directly in a browser. If you go to the tests
> > directory in each major project (like dojo or dijit), you should see a
> > runTests.html. Open that in the browser of your choice, and DOH will
> > run the tests in that browser.
>
> Yes, certainly I've done this. But for continuous integration, there's
> certainly value in being able to run the tests from command-line, even if
> they consists largely of visual tests. It gives you some protection against
> regression errors and catches newly introduced run-time errors. My interest
> in this is to evaluate DOH's suitability for testing our own Javascript code
> that uses dojo. But it seems DOH is geared towards testing dijit and
> dojo-based widgets rather than for general-purpose testing framework. And
> that there's a fair bit of work to do if you want to integrate DOH with
> larger build systems following continuous integration principles.
Yes, if you want something that can be triggered from the command line
and catch the results on the command line, then DOH is not designed
for that purpose. I find that this is an area that is lacking for
browser development, but it is also a tricky problem since each
browser engine has different quirks.
> > The tests that run in rhino are the non-browser tests (normally the
> > pure JS tests). There may be a reporting issue with the number of
> > tests (I have not confirmed how it calculates the numbers or how it
> > reports what tests are run), but while it does use the same master
> > test file for Dojo core (dojo/tests/module.js), some tests are
> > excluded by checking to see if dojo.isBrowser is true. It will be
> > false for the rhino case, so those tests will be skipped. Also note
> > that while it looks like some tests in dojo/tests/module.js seem to be
> > included for the rhino enviroment, if you look inside the specific
> > tests being dojo.required, they may do the dojo.isBrowser check within
> > that test file.
>
> Ok, so that seems to be what's happening. And it doesn't really say which
> tests are run, just "GROUP XXX has 10 tests to run" and in the end, the
> number of all tests and if there were any errors in them, but nothing about
> how many tests were actually run. In rhino, it also completes much quicker
> than in browser via the runTests.html which implies most of the tests were
> skipped in the former.
>
> Crosscheck, which I mentioned previously, seems to be about the only
> browser-emulation environment for Javascript testing. Have you or anybody at
> or affiliated with the dojo team considered using it instead of DOH or
> possibly merging the two?
I believe it is more important that we can prove the Dojo code runs
correctly in each browser engine, and trying to rely on an emulation
framework will always be suboptimal, since that emulation framework
cannot capture all of the nuances of each browser engine, and it will
usually lag behind the actual browser engine development (case in
point, Crosscheck only advertises being able to test against IE 6,
Firefox 1.0 and 1.5).
I would like to see a command-line version of a testing tool, but I
expect it to take a form of running the real browser engine, putting
hooks in that engine to collect test results, and then relay those
results to the command line. That would be neat if someone built that,
but it would require a specific wrapper for each browser engine
(probably in C/C++), and to get full test coverage, you would need to
run it on multiple operating systems (since IE 6 for example does not
run on linux or OS X).
James
_______________________________________________
FAQ: http://dojotoolkit.org/support/faq
Book: http://dojotoolkit.org/docs/book
Forums: http://dojotoolkit.org/forum
Dojo-interest@...
http://dojotoolkit.org/mailman/listinfo/dojo-interest
> On 12/24/07, James Burke <jburke@...> wrote:
>
> > The built version that you get from the 1.0.2 download location (not
> > the -src file) is built for the browser, and not for rhino. Running
> > that version with rhino and DOH can produce the "navigator" issue.
>
>
> Thanks a lot James for your answers, this is really helpful. Quickly looking
> at the build.xml, I didn't see a target that would specifically "build for
> the browser" and I couldn't find any documentation about it. What target
> should I run to create this version? And why is it needed - wasn't one of
> the design goals of dojo 1.0 to target purely a browser environment?
The default behavior of the build system is to build for the browser.
Looking more at it now, the build system has code in it to build a
rhino-specific build, but that build process fails right now. So usage
in rhino right now means just using a source distribution. I filed
this trac bug for doing a build for the rhino environment (I just
fixed it in the repository trunk):
http://trac.dojotoolkit.org/ticket/5467
See that ticket for an example profile file that makes a build for the
rhino environment (requires the latest trunk code to work).
> > Dojo can be run in the browser or in rhino, but the built versions we
> > put on the website are only for running in the browser. We do not
> > support auto-detection of the environment in the built version to cut
> > down on the size of dojo.js and the vast majority of users only care
> > about the browser version. So, for running in rhino, you will need to
> > get the -src download or grab the source from subversion.
>
> It'd be good to document this at
> http://dojotoolkit.org/book/dojo-book-0-9/part-4-meta-dojo/d-o-h-unit-testing
> (or presumably upcoming dojo-book-1-0).
I put a note near the Rhino testing section about using a source
distribution when running DOH in Rhino.
> > DOH does not provide a simulation of the browser environment in rhino.
> > Instead, it runs directly in a browser. If you go to the tests
> > directory in each major project (like dojo or dijit), you should see a
> > runTests.html. Open that in the browser of your choice, and DOH will
> > run the tests in that browser.
>
> Yes, certainly I've done this. But for continuous integration, there's
> certainly value in being able to run the tests from command-line, even if
> they consists largely of visual tests. It gives you some protection against
> regression errors and catches newly introduced run-time errors. My interest
> in this is to evaluate DOH's suitability for testing our own Javascript code
> that uses dojo. But it seems DOH is geared towards testing dijit and
> dojo-based widgets rather than for general-purpose testing framework. And
> that there's a fair bit of work to do if you want to integrate DOH with
> larger build systems following continuous integration principles.
Yes, if you want something that can be triggered from the command line
and catch the results on the command line, then DOH is not designed
for that purpose. I find that this is an area that is lacking for
browser development, but it is also a tricky problem since each
browser engine has different quirks.
> > The tests that run in rhino are the non-browser tests (normally the
> > pure JS tests). There may be a reporting issue with the number of
> > tests (I have not confirmed how it calculates the numbers or how it
> > reports what tests are run), but while it does use the same master
> > test file for Dojo core (dojo/tests/module.js), some tests are
> > excluded by checking to see if dojo.isBrowser is true. It will be
> > false for the rhino case, so those tests will be skipped. Also note
> > that while it looks like some tests in dojo/tests/module.js seem to be
> > included for the rhino enviroment, if you look inside the specific
> > tests being dojo.required, they may do the dojo.isBrowser check within
> > that test file.
>
> Ok, so that seems to be what's happening. And it doesn't really say which
> tests are run, just "GROUP XXX has 10 tests to run" and in the end, the
> number of all tests and if there were any errors in them, but nothing about
> how many tests were actually run. In rhino, it also completes much quicker
> than in browser via the runTests.html which implies most of the tests were
> skipped in the former.
>
> Crosscheck, which I mentioned previously, seems to be about the only
> browser-emulation environment for Javascript testing. Have you or anybody at
> or affiliated with the dojo team considered using it instead of DOH or
> possibly merging the two?
I believe it is more important that we can prove the Dojo code runs
correctly in each browser engine, and trying to rely on an emulation
framework will always be suboptimal, since that emulation framework
cannot capture all of the nuances of each browser engine, and it will
usually lag behind the actual browser engine development (case in
point, Crosscheck only advertises being able to test against IE 6,
Firefox 1.0 and 1.5).
I would like to see a command-line version of a testing tool, but I
expect it to take a form of running the real browser engine, putting
hooks in that engine to collect test results, and then relay those
results to the command line. That would be neat if someone built that,
but it would require a specific wrapper for each browser engine
(probably in C/C++), and to get full test coverage, you would need to
run it on multiple operating systems (since IE 6 for example does not
run on linux or OS X).
James
_______________________________________________
FAQ: http://dojotoolkit.org/support/faq
Book: http://dojotoolkit.org/docs/book
Forums: http://dojotoolkit.org/forum
Dojo-interest@...
http://dojotoolkit.org/mailman/listinfo/dojo-interest
« Return to Thread: [1.02] ReferenceError: "navigator" running tests in Rhino
| Free Forum Powered by Nabble | Forum Help |