On Thu, Jan 3, 2013 at 2:33 PM, Andrew Sutherland
<[email protected]> wrote:
> The e-mail (packaged) app does an asynchronous XHR to check whether there is
> an autoconfig file in its local /autoconfig/ directory before moving on to
> asking the internet for answers.  So if you type in "[email protected]" for your
> e-mail address, we will do an XHR for "/autoconfig/yahoo.es".

And you are doing this from a page with a URL that looks something like

app://email.gaiamobile.org/index.html

Is that correct? I.e. you are doing an XHR request to something like

app://email.gaiamobile.org/autoconfig/yahoo.es

> Previously, our onload() event would fire with the failure when a match was
> not found.  However, it now appears that in the nsXMLHttpRequest::Send call,
> that the mChannel->AsyncOpen is newly failing with NS_ERROR_FILE_NOT_FOUND.
> This ends up being thrown as an exception:
> [Exception... "File error: Not found"  nsresult: "0x80520012
> (NS_ERROR_FILE_NOT_FOUND)"  location: "JS frame ::
> app://email.gaiamobile.org/js/ext/gaia-email-opt.js :: getXmlConfig :: line
> 35948"  data: no]

Is this only happening if the "autoconfig/yahoo.es" file doesn't exist
in the app package? I.e. is the load completing successfully if the
file exists.

> From https://bugzilla.mozilla.org/show_bug.cgi?id=804395, it seems like we
> want XHR to not change in behaviour between when something is run as a
> packaged app versus when it is run on the web.  So this seems undesirable.
> This may also be a spec violation since what is happening only seems
> suitable for a synchronous XHR, although I have not done a deep read of the
> XHR spec lately and only skimmed it today.
>
> My desired outcome from this e-mail is if someone could indicate whether the
> exhibited behaviour is desired,

I agree that throwing from the send() call like that is not desired.
And I think that it is a violation of the XHR spec.

> if we think this changed recently, and if
> anyone has an idea of what bug regressed this if so.  (I've done some quick
> scanning, but there are many bugs touching the general area of code, but
> none are jumping out at me very obviously.)
>
> https://bugzilla.mozilla.org/show_bug.cgi?id=826468 is currently tracking
> the e-mail problem.  We aren't really sure when this might have regressed
> right now.

A guess is that this is a regression from bug 815523.

If that is indeed the case, mark bug 826468 as blocking bug 815523 and
nominate it for for blocking-basecamp.

But I am worried that if this is too complicated to fix (hard to say
without knowing what's broken) then we might not be able to fix it in
time for launch.

/ Jonas
_______________________________________________
dev-b2g mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-b2g

Reply via email to