flex-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Michael Schmalle <teotigraphix...@gmail.com>
Subject Re: [FALCON] Definitely needs some love ...
Date Fri, 18 Mar 2016 19:16:58 GMT
Yes and at the complexity of the transpiler and amount of people that have
actually touched the code, it's all understandable.

When I started that project, I actually used the strategy of unit testing
pretty much every AS statement and expression to produce valid AS, I didn't
even think about JS production in the beginning.

Once the GOOG emitter was started and things got complex, the standard
100's of unit tests I wrote became mute.

IMHO I think th unit tests right now are a mess only due to the fact that
if you realized something was wrong in the impl it might be an actual pain
in the ass to fix it because there is so many things dependent on each

And yeah, there is not a lot of course correction in the logic and it was
brute force from the beginning to get something that actually was close to


On Fri, Mar 18, 2016 at 3:06 PM, Christofer Dutz <christofer.dutz@c-ware.de>

> Hi Josh,
> the problem is that sometimes I can't even say what's the cause of the
> npes. Most of them occured inside the compiler and from the context
> variables I can only guess what's going on. And I would probably have to
> open a big 2 digit number of issues, I would like to prevent that. I guess
> as soon as I have ported things to maven I can start using some of my
> static code analysis tools and start mass-producing test-cases ;-)
> But I would like to suggest to start thinking about the case that people
> do things wrong. FlexJS works well if you are on the correct path, but the
> code currently simply can't cope well with errors.
> Chris
> ________________________________________
> Von: Josh Tynjala <joshtynjala@gmail.com>
> Gesendet: Freitag, 18. März 2016 15:59
> An: dev@flex.apache.org
> Betreff: Re: [FALCON] Definitely needs some love ...
> I think the first step is documenting these null pointer exceptions as they
> are encountered. Opening bug reports in JIRA seems like the right course of
> action.
> - Josh
> On Fri, Mar 18, 2016 at 7:09 AM, Christofer Dutz <
> christofer.dutz@c-ware.de>
> wrote:
> > Hi,
> >
> >
> > Well I think one thing we definitely need to work on, is making Falcon
> and
> > FalconJX a lot more robust. During my work on Mavnizing everything I
> > continuously encounter NullPointerExceptions all over the place. It seems
> > that most of the code seems to currently work if everything is correct,
> but
> > fail miserably with NullPointers if you are not 100% correct. Guess this
> > will cause a lot of frustration with our users.
> >
> >
> > Chris
> >

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message