[Marble-devel] Tessellation

[bbeschow at cs.tu-berlin.de]

> Hi,

Hi Thibaut,

first of all thanks for working on this complex beast of code!

> While solving the Idl issues I had tried to move the tessellation out of
> the way.
> The solution proposed in the review moved it in the LineString.
> I am coming with the following goods and bads:
>
> in the bads:
> - it will not be possible to have screen point density specific per
> projection.
> - tessellation "sounds" like a projection issue, and a solution in the
> LineString provides a generic solution which is artificial. If I remember
> some xkb-style projection trivia, some specific projection in some
> even specific cases might not need tessellation at all... (i think
> stereoscopic orthodromic paths going through the center of projection,
> and spherical orthodromic paths going through the center of the screen)
>
> in the goods:
> - i didn't see regressions in the visual display. Maybe there are more
> (and too many) points sometimes.
> - it totally removes idl considerations which spherical projection has
> no interest in, and which cylindrical projections solve independently
> later on.
> - for a given zoom, it should be possible to cache this tessellated
> geodetic calculus, which is relatively costly as benchmarked with
> callgrind in the (not very representative) test cases. Therefore I expect
> panning would benefit from this.
>
> I prepared a new patchset this week-end, leaving tessellation with
> abstract projection  and fixing the idl issues still.
>
> I propose to close the first review, directly push whatever fixes do not
>  depend on either tessellation in the LineString or in the Projection,
> and discuss more the tessellation part.

If the review request is obsolete by now I think it's a good idea to close
it in order to avoid people potentially spending time on it.

> Do you already have an opinion on this issue, or should i propose the 2
> patches, or should we have an agreement beforehand?

Having your fixes in master ASAP sounds like a good idea to me. However,
since the code you're touching here is one of the more complex ones inside
Marble, you might want to have it tested by some fellows. I wouldn't mind
though if you just pushed your changes if you put your fixes under unit
test and be responsive if regressions are observed.

Moreover, I'm interested about the problems your patch will fix. In
particular, I filed bug 315922 regarding tessellation and I wonder whether
your patch already fixes it or whether I should create a new review
request.

Greetings,
Bernhard