http://bugs.winehq.org/show_bug.cgi?id=12730
--- Comment #7 from Jörg Höhle hoehle@users.sourceforge.net 2009-03-11 08:18:44 --- Created an attachment (id=19881) --> (http://bugs.winehq.org/attachment.cgi?id=19881) excerpts from winetest version 1.1.15 logfiles in 16 vs. 32bpp
Attached is a summary of differences between running winetest in 32bpp and 16bpp modes on the same machine.
Tests from build 8e39646ac4d256b813ff69a23ba0c62ab4f08d33 (wine-1.1.15)
Differences are in: - visual.c - 3 different files?! (one per directx dir?) - bitmap.c - GetDIBits; also performs significantly less tests - palette.c - getColor - input.c - msg.c - win.c - GetForegroundWindow
o In 16bpp the three visual.c tests are mostly skipped, although the summary doesn't say so: visual.c:1306: Sanity check returned an incorrect color(00000000), can't assure the correctness of the tests, skipping visual: 3 tests executed (0 marked as todo, 0 failures), 0 skipped. (Likewise on http://test.winehq.org, some entries are green because almost no test was performed). In 32bpp, plenty of tests would be executed in visual.c.
o Some differences in the number of tests executed are not understandable, as there's no mention of skipping anything nor any difference in the log file, e.g.: sock: 13181 tests executed (5 marked as todo, 0 failures), 0 skipped. vs 32bpp: sock: 13899 tests executed (5 marked as todo, 0 failures), 0 skipped.
o Noteworthy are tests where there are more errors (or only errors) in 32bpp: - editor.c:4590: Test failed: test paste: strcmp = 1, text='testing paste Ts' - menu.c:1826: Test failed: test 0, 2, 4, 6, 8, 10, 12, 14, 16 - msg.c: sequence; testing xyz press/release
Finally, msg.c performed more tests and better results in 16bpp! msg: 13057 tests executed (52 marked as todo, 24 failures), 2 skipped. vs. with 16bpp: msg: 14296 tests executed (52 marked as todo, 2 failures), 2 skipped. Perhaps msg.c was written in 16bpp times?
In 1.1.15, "Meine Traumburg" still exhibits the 16bpp bug I mentioned in comment #3. (I promised to open a distinct issue should 16bpp tests become as successful as 32bpp while this bug persists)