[cla] dojo/on.js : fixTouchListener is undefined

[loodwig]

in dojo/on.js, the addListener() function attempts to correct the listener if has("touch") is true. However, within the addListener() function is no reference to fixTouchListener(), and an error is thrown / execution is halted when the user agent is:

Mozilla/5.0 (Linux; U; Android 2.3.6; en-us; Nexus S Build/GRK39F) AppleWebKit?/533.1 (KHTML, like Gecko) Version/4.0 Mobile Safari/533.1

The fixTouchListener() function appears to be defined later in the file, but it is not in scope of its calling components at line 120 and line 127. To solve this issue, I recommend the function be brought within the scope of addListener(), preferably within the if(has("touch")){ block, so that it is defined when needed for mobile browsers that support touch.

I have attached a diff from 27684 (HEAD at the time of making this ticket), which shows my proposed solution.

[loodwig]

diff between 27684 and my suggested solution

[Colin Snover]

Have you signed a CLA or are you under a CCLA? If so, what is your name? If not, could you please do so at ​http://dojofoundation.org/about/cla? Thanks!

[loodwig]

Replying to csnover: CLA form is being faxed now.

[loodwig]

html and js needed to reproduce the problem

[Colin Snover]

CLA received. Thank you!

[Colin Snover]

Kris can you review/triage?

[Kris Zyp]

I don't follow this ticket. fixTouchListener is defined in the module's factory scope, which contains addListener and can be referenced by the addListener. This function is being successfully called on touch events, and touch events are being properly normalized by this function. I can't see anything happening wrong (or anything at all) in the attached test case.

[loodwig]

You're absolutely right, I've only been able to get this to happen using the 1.7.1-src unzipped locally. If I use trunk, the cdn, or the download of 1.7.1 (built), this problem goes away. I'll continue to research, as the problem appears to only be on my end. Sorry for that.

[loodwig]

Replying to kzyp: Hmm, it appears that with further testing I've ruled out my environment to the best of my knowledge. I've added the following site that demonstrates (using the attached code) the error consistently with the aforementioned android user agent. I'm noticing it 100% of the time using chrome set to the user agent as well as a android device. I figured a public site might help:

​http://loodwig.org/14720/

[Colin Snover]

That test case does not raise an error, just says “Loading…”. What is it supposed to do? What are the reproduction instructions?

[Kenneth G. Franqueiro]

I've also attempted to reproduce an error locally using the attached reproduction material, running Android 2.3.7 with weinre loaded in the page, but I see no errors.

Based on what I heard discussed in the IRC channel earlier, it sounds like the error was only reproduced when UA-spoofing was used to attempt to test the page in an "Android-like" environment but from a desktop browser - in which case, any has("touch") code won't be executed, which in this case includes the code that both calls and defines the function in question.

Either way, I'm not sure how this error could be coming up... I agree with Kris' assessment.

[loodwig]

I concur as well. I've tested on a few other android devices outside of my testing environment, and I am not getting the aforementioned error. Initially, I was getting a different but similar error (which may very well be with my code). At this point, I'll assume all errors I've received came from my "spoofing" of the user agent in order to create a test case to diagnose this separate error. If (and I emphasize if) there is an error with dojo, it's a separate issue. In the future, I'll do better testing before filing a ticket.