1)
We started with RTT 1.8.x so maybe you are right and that issue is already
fixed

2)
It will be great if we were able to set the precision in an environment
variable like ORO_DOUBLE_PRECISION = 0.6. But for now

3)
This patch is more problematic since, it will cut the “Backward compatible”
issue of RTT 1.x . The problem comes in the implementation of ACE&TAO who
throws and catches CORBA::BAD_KIND exception when using Sequence<double>
only. Since we are using MSVC (in the output section), we can see a lot of
these false problems exceptions, and that’s hide the real exceptions that we
would like to take care.

4)
Well, segmentation fault occur there we MSVC, so if you don’t see difference
in the result apply that patch for Win32 users. Actually, we can produce the
problems when browsing (with the ctaskbrowser) a taskcontext (with
attributes) built with rtt 1.8.x-msvc-patched.

I can easily understand that you don’t agree with the issue 3) because it
breaks backward compatibility. The patch 2), can be seen as temporary.
However, it will be great if 1),4) can be added in the next RTT versions.

Thanks for the answers.

Simon

On Mon, Feb 8, 2010 at 5:59 AM, Peter Soetens <peter [dot] soetens [..] ...>wrote:

> On Friday 05 February 2010 16:58:14 S Roderick wrote:
> > On Feb 5, 2010, at 10:40 , Simon Pelletier-Thibault wrote:
> > > Hi,
> > >
> > > Our compagny work with Orocos RTT and we had make some patches.These
> > > patches were useful in our application and we would like you to
> introduce
> > > them in the next version of Orocos. Since we versionize it on our SVN I
> > > give you the diff we made.
> > >
> > > We did some patches on the RTT. These patches were useful in our
> > > application and we would like you to introduce them in the next version
> > > of Orocos. Since we versionize our own version on our SVN I give you
> the
> > > diff we made.
> > >
> ...
> > > May you apply the changes in the next release of Orocos 1.x?
> > > May you give me feedbacks?
> > >
> > > Simon Pelletier-Thibault
> >
> > Simon, thanks for the submitting the patches. While Peter is the
> > authoritative person here, I can give you some feedback on these.
> >
> > 1) I think this is an issue Peter solved in trunk a while ago.
>
> We had issues with processing nameless properties in xml file, there was
> also
> an issue with saving properties. Both were indeed fixed and backported to
> 1.x.
> So probably not needed.
>
> >
> > 2) We do something like this also, but we can't put that in trunk and
> > expect everyone to want the same precision in output. The typing system
> > needs an overhaul to allow formatting specifies somewhere along the
> line.
>
> I have not spent any time on this either, not easily fixable. This patch
> will
> not get in.
>
> >
> > 3) There is a double-up in "DoubleSequence" between RTT's plugin and
> KDL's
> > plugin, IIRC. I think I put a patch in for this a while back, but Ruben
> > (KDL) has been busy ...
>
> "Transporting vector<double> gives Bad kind exception in debuging with
> MSVC,
> This doesn't give exceptions"
>
> That's the patch I was least certain about of what it actually fixes, or to
> put
> it differently, what needed to be fixed ? Do you have more details about
> what
> went wrong on which side of the original code ?
>
> >
> > 4) I don't see the differences here
>
> Patch is fine for me if it helps Win32 users.
>
> Peter
>
>