[Testing] Google Test Automation Conference & OLPC
Samuel Greenfeld
greenfeld at laptop.org
Fri Nov 4 23:28:18 EDT 2011
Last week I was at the Google Test Automation Conference (GTAC), an annual
event held this year in Mountain View, California, USA.
The videos and source material from this conference will be posted to
http://www.gtac.biz/talks , with most of the videos already online. Past
years' conferences may also be found on YouTube.
There were a variety of talks this year on web and browser automation, as
well as on other things. But from an OLPC perspective I just wanted to
touch on several talks that I personally considered relevant, sorted by
their appearance in the conference as well as on the aforementioned web
page:
- "Test Is Dead" (Alberto Savoia): The opening keynote focused on how
companies in many fields are getting rid of software testers and holding
developers more accountable for their code. The focus now is "making the
right it, not making it right" to verify that concepts are viable to a wide
audience as quickly as possible.
The longer a company spends on a development path, the harder it is to
change it later. Rapidly releasing new versions can help find "idea bugs"
faster while leaving "code bugs" users tolerate.
From an OLPC/Sugar perspective, we should try and do this with user
interface changes, etc., including getting feedback from deployments as
well other end-users. But it may not be possible for us to completely
eliminate my job, as many deployments only update their XO software once
per year or so.
This may be important for verifying the Nell[*] proof-of-concept as
something which schools/foundations/countries/governments/etc. will want to
pay for as opposed to something which just sounds cool to do.
[*] http://cananian.livejournal.com/65077.html
- Developing and Testing WebGL (Roy Williams): This may also be relevant
to Nell. WebGL (a 3D programming language) is usually rendered directly by
the graphics card. This leaves the WebGL canvas appearing more or less as
a "black box" inside the browser. It is therefore somewhat hard to
automate WebGL testing, as there is no guarantee that you can see what is
inside of it.
A Model/View/Controller programming style is recommended to make
debugging and testing of WebGL applications easier. Google also released a
tool to help do WebGL testing at this event (one of many tools whose source
code was released at this conference), but I have not had a chance to look
into its abilities.
- BidiChecker: Automated Bidi Testing of Web Applications (Yana Margolin
& Jason Elbaum): Not every language is written left-to-right. Arabic,
Hebrew, and some others are written right-to-left.
But within right-to-left languages, some items (numbers, etc.) may still
be written left-to-right. It may also be desirable to mix right-to-left
and left-to-right languages within the same interface.
When testing translations, it is not just important to make sure that
the insertion order of parameters is correct for each translation within a
single language, but also the text direction of each translated string &
its parameters. Checks also need to be done on combinations of
translations as well as the entire layout of an interface as a whole.
While I believe GTK+ tries to automate much of this for us, there is no
guarantee that its default approach will be correct. We therefore need
native and/or well-taught speakers to tell us if something looks incorrect.
- How to Hit a Moving Target in the Cloud (Vishal Chowdhary): This talk
focused on verifying machine-based language translation, where the
"correct" result output for a translated sentence could be one of many
results.
One therefore has to look at pieces of the translation puzzle to figure
out if the output makes sense without re-writing your own translation
engine. Having people test and provide feedback also is useful. This
results in gray answers for test results, but can give you an idea of if
something is terribly incorrect.
This also may be relevant to Nell, where the user's progress may control
how the environment changes.
- How Hackers See Bugs (Hugh Thompson): The first day closing keynote
does not talk about "bugs" in the traditional software sense, but rather
covers a variety of issues. It looks at how trust relationships can be
exploited, and how security measures and public data from past years can be
exploited to wreck havoc in the present.
This talk also contains a few examples of why it important to present an
interface that users can intuitively understand how to use, especially when
they are not given clear instructions on how to use it. Such an approach
is generally relevant to Sugar.
- Latest in Google Test Tools (Ibrahim El-Far & Jason Arbon): Mostly
interesting from a manual test perspective, as Google has taken out a lot
of the side tasks related to manual testing. I think it may be worthwhile
to write a Sugar activity which does similar.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.laptop.org/pipermail/testing/attachments/20111104/1af2aaa3/attachment.html>
More information about the Testing
mailing list