[OSM-dev] [OSM-talk] Osmarender street names rendering regression

[Dirk-Lüder Kreie]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[added some old quotes for CC to dev ML]

Dirk-Lüder Kreie schrieb:
> Ben Laenen schrieb:
>> Dirk-Lüder Kreie schrieb:
>>> Dirk-Lüder Kreie schrieb:
>>>> Ben Laenen schrieb:
>>>>> Hi,
>>>>>
>>>>> Since a few days there seems to be a regression in the rendering of
>>>>> street names in Osmarender. The text is badly positioned and overlaps
>>>>> itself (like upside down right to left, and normal left to right again
>>>>> at the same spot). The tags are good, the direction of the segments is
>>>>> good, as well as the order of the segments in the way. This also didn't
>>>>> happen before. Some other names are suddenly placed upside down as
>>>>> well.
>>>>>
>>>>> There are a few examples visible here:
>>>>> http://www.informationfreeway.org/?lat=6653882.34077&lon=493071.27604&zoom=16&layers=B00
>>>>> They showed up the latest four times I requested a re-rendering of that
>>>>> area, so it shouldn't be just one renderer behaving badly.
>>>>>
>>>>> Could it be that this problem is related to the automatic correction of
>>>>> segment directions and orders in ways? As it started happening around
>>>>> the same time those corrections were being applied before rendering.
>>>>>
>>>> I just rendered that area with my frollo-enabled client and it comes out
>>>> clean.
>>>>
>>>> I will try without frollo just to see what happens.
>>>>
>>> What you saw is the output without frollo so those ways in Antwerpen are no longer ordered, for whatever reason.

>> Things I think cannot happen -- even with completely wrong ordering --
>> are:
>
>> * ways that consist of one segment that have the street names upside
>> down
>
> Osmarender renders the names of right-to-left-streets upside down, only
> a test detects this and reverses the text direction, however this test
> fails if the segments in the way are not ordered, or point in different
> directions.
>
>> * the street names that render in two directions on the same segment
>
> This can happen if there are 2 copies of the way -- one ordered, one
> unordered

Apparently there was indeed a regression because osmarender seems to
*depend* on frollo to figure out right-to-left-streets correctly. since
a frollo failure is not considered critical, tiles at home renders the
streets without frollo but produces the output you saw, where a
Cambridge client would have produced nice output.

CC to dev

Dirk-Lüder "Deelkar" Kreie


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGBbf/FUbODdpRVDwRAqz2AJ0ar7kCwUGljs+nQkRM2fLPwEhoGwCgxpTr
+Eqs7jELUzFk5rTmLxdMzY4=
=ROVk
-----END PGP SIGNATURE-----