Font rendering with transparency in PNGs using PC256

[posthumusb]

Version: MapServer 6.0.1 (from MS4W 3.0.1) OS: Windows XP Pro SP3

In MapServer 6.0, output PNGs using IMAGEMODE PC256 and TRANSPARENT ON in the output format results in corrupted text when rendering truetype fonts. This worked fine in MS 5.4 and 5.6.

Here is the output format causing this issue:

OUTPUTFORMAT
  NAME png256
  DRIVER "GD/PNG"
  MIMETYPE "image/png"
  IMAGEMODE PC256
  EXTENSION "png"
  TRANSPARENT ON
END

Attached is a stand-alone mapfile with a truetype font and font-list file. Resulting PNGs using MS 5.4 and MS 6.0 are also included for comparison.

This issue was first raised in this post: ​http://osgeo-org.1803224.n2.nabble.com/Font-rendering-in-MapServer-6-0-tt6555935.html

Though using the AGG renderer is an option, it creates new challenges when working with Internet Explorer 6 (see ​http://osgeo-org.1803224.n2.nabble.com/Rendering-with-a-single-transparent-colours-in-MapServer-6-tt6769109.html). Having a fully functional PC256 option would alleviate this problem.

[posthumusb]

Mapfile showing font rendering bug using PC256 with transparency

[dmorissette]

Rendering issue. Reassigned to tbonfort.

[posthumusb]

Here's a request to go along with the mapfile:

​http://localhost/cgi-bin/mapserv.exe?map=/tmp/test_pc256.map&layers=point&styles=&service=WMS&width=400&height=300&format=image/png&request=GetMap&srs=EPSG:26913&version=1.1.1&bbox=460000,5500000,600000,5700000

[tbonfort]

ccing Steve as he's the GD expert

[posthumusb]

mapgd.c patch for ticket 4013

[posthumusb]

In mapgd.c, there's a condition that converts the background colour to purple (117,17,91) if either the background colour does not exist, is not valid, OR transparency is turned on. This purple colour is bleeding into the text around the edges, causing the jagged appearance of the text.

I've attached is a potential patch that sets the background to (117,17,91) only if the background colour does not exist or is invalid, and then worries about transparency later. This fixes the issue, though I can't say if it breaks something that maybe was relying on the purple background (hopefully not).

[posthumusb]

Steve or Thomas, can either of you have a spare moment, can you take a look at the patch submitted? It's a small fix that has been working well for me over the last month or so, but maybe it doesn't work for all scenarios that one of you would know of.