Large number of failed tests in autotest suite after merge of mapserver6 sandbox (rendering api) into trunk

[rouault]

Just to make sure MapServer 6 doesn't get released without a pass on this...

I've observed the following results in msautotest/wxs with latest trunk (r10806): 18 tested skipped 134 tests succeeded 114 tests failed

I see similar pattern on Tamas daily builds ( see ​http://vbkto.dyndns.org:1280/sdk/build-output/vc7-20101218-4-18-59-67-vc7-dev.txt for example )

I've bissected it to be due to r10743 :

r10742 : 18 tested skipped 227 tests succeeded 21 tests failed

r10743 : 18 tested skipped 146 tests succeeded 102 tests failed

[sdlime]

Test Assessment (r10829)

[sdlime]

I started through the misc tests. I fixed the mapfiles that needed updated IMAGETYPEs (r10829). I looked at each of the tests (misc only) and made an initial assessment. Many suffer from slight image rendering differences and probably should just have new expected results created. There are a number of bugs though (see spreadsheet)...

Note: Tests were run on MacOS with trunk (r19829) with the following version string "MapServer version 5.7-dev OUTPUT=GIF OUTPUT=PNG OUTPUT=WBMP OUTPUT=SVG SUPPORTS=PROJ SUPPORTS=AGG SUPPORTS=FREETYPE SUPPORTS=ICONV SUPPORTS=WMS_SERVER SUPPORTS=GEOS INPUT=POSTGIS INPUT=OGR INPUT=GDAL INPUT=SHAPEFILE". I can't seem to get jpeg enabled which is why those tests weren't run.

Now on to GDAL tests.

Steve

[tbonfort]

I'll have a stab at the tests to update the expected files when they are ok. I won't be able to do so before around the 15th of january though.

thomas

[tbonfort]

By the way, thanks Steve for setting up the spreadsheet, must have been a tedious task !

I've published the speadsheet as a google doc so we can collaboratively update it: ​https://spreadsheets.google.com/ccc?key=0AnCiCdIXpTHidHFYR3VKVXdRS3JQd2cydmFENk5Gcmc&hl=en&authkey=CLr-pf8H

Although there is a high number of fails, only a few of them are significant.

• all the differences related to text are to be expected, as there have been significant changes here.
• there's a problem with spacing on text that should be delt with. in essence, the pixel coordinates for text placement must be rounded so that the font engines can do correct hinting (we end up with blurred text otherwise), but this should not be done for rotated (i.e. angle follow) text, thus the funky character placement on the angle follow tests.
• changes due to bitmap font sizes and weight cannot be corrected, unless someone wants to create bitmap fonts for agg that are strictly identical to the ones provided by gd.
• vector symbols in previous versions where uncorrectly rotated. this is now corrected thus the significant difference. will have to check on the scaling though.

[tbonfort]

the tests have been updated, and individual failures should now be dealt with. this bug can now be closed as we are not facing widespread failures now.