<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 01/23/2014 06:19 PM,
<a class="moz-txt-link-abbreviated" href="mailto:firefox-dev-request@mozilla.org">firefox-dev-request@mozilla.org</a> wrote:<br>
</div>
<blockquote
cite="mid:mailman.42141.1390497587.24863.firefox-dev@mozilla.org"
type="cite">
<pre wrap="">Message: 8
Date: Thu, 23 Jan 2014 12:19:24 -0500
From: Benjamin Smedberg <a moz-do-not-send="true" class="moz-txt-link-rfc2396E" href="mailto:benjamin@smedbergs.us"><benjamin@smedbergs.us></a>
To: Firefox Dev <a moz-do-not-send="true" class="moz-txt-link-rfc2396E" href="mailto:firefox-dev@mozilla.org"><firefox-dev@mozilla.org></a>, Myk Melez <a moz-do-not-send="true" class="moz-txt-link-rfc2396E" href="mailto:myk@mozilla.com"><myk@mozilla.com></a>
Subject: What is the status of Webapp installation in Firefox?
Message-ID: <a moz-do-not-send="true" class="moz-txt-link-rfc2396E" href="mailto:52E14F1C.20409@smedbergs.us"><52E14F1C.20409@smedbergs.us></a>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
I am a bit concerned about the situation with the situation of webapprt
in Firefox desktop. I discovered recently that webapp installation is on
by default, at least in Nightly. Is that true on all release channels? I
also discovered, when I tried to install an app linked from twitter,
that webapps wouldn't start on Windows because of some ICU linking
changes. That was filed and now fixed in bug 958108, but it seems to
indicate that we don't have any automated testing of the webapp system
at all, which is very concerning to me.
The last I checked, the web app system in desktop builds was considered
a labs/experiment. Is that still the case? If so, should we turn it off
by default until we have decided that this is a feature we want to ship,
and we're confident in its engineering/automated testing?
As part of looking into this, I discovered that webapprt crashes appear
to be submitted as Firefox crashes and there isn't a simple way to
differentiate them: bug 962163 is actually filed as a <b class="moz-txt-star"><span class="moz-txt-tag">*</span>topcrash<span class="moz-txt-tag">*</span></b> for
Firefox nightly even though I'm about 99.9% certain that it is all
webapp launch crashes. And the stack traces and symbols for these webapp
crashes are broken; they report the launch module as "yXpJpNW.TQO"
without a debug ID or symbols; I'm pretty certain that this is actually
webapprt-stub.exe with minimal repackaging to get an icon embedded in
the .exe.
--BDS</pre>
</blockquote>
<br>
It's been enabled in release versions for more than 10 versions now.<br>
<br>
For installations, I'm adding new tests in bug 898647 (that
unfortunately can't run on Mac), but<br>
we already have some basic tests.<br>
<br>
The tests for the webapp runtime (that actually run it) are in our
codebase, but aren't executed automatically,<br>
there's a bug I filed some months ago to automatically run them (bug
899707 and bug 920569).<br>
With these enabled I think we'd have a good coverage.<br>
<br>
- Marco.<br>
</body>
</html>