January 2010 - Wine-devel - list.winehq.org

Re: [PATCH 1/2] msi: queue dynamically allocated strings in cond.y
by James Hawkins 05 Jan '10

05 Jan '10

On Sat, Jan 2, 2010 at 10:36 AM, Nathan Gallaher <ngallaher(a)deepthought.org> wrote: > > +struct cond_mem { + struct list entry; + void *ptr; +}; + +static void cond_free( void *info, void *ptr ) +{ + COND_input *cond = (COND_input*) info; + struct cond_mem *mem, *safety; + + LIST_FOR_EACH_ENTRY_SAFE( mem, safety, &cond->mem, struct cond_mem, entry ) + { + if( mem->ptr == ptr ) + { + msi_free( mem->ptr ); + list_remove( &(mem->entry) ); + msi_free( mem ); + return; + } + } + ERR("Error freeing %p\n", ptr); +} This won't fly. cond_free needs to be an O(1) operation, like your original patch. -- James Hawkins

2 4

Re: [tools]winetest: fix child-result coloring and counting
by Paul Vriens 05 Jan '10

05 Jan '10

On 01/04/2010 08:50 PM, André Hentschel wrote: > You can see wrong coloring e.g. here http://test.winehq.org/data/f74e312bf81032c46bd2ce080a5f6a33c7ac3ee3/wine_a… > Also the counting is wrong, that unit test is shown in the summary with 2 errors, but only has 1 > This patch fixes the issue. The totalize of those variables is already made in the "offline" software > --- > winetest/dissect | 6 +++--- > 1 files changed, 3 insertions(+), 3 deletions(-) > > diff --git a/winetest/dissect b/winetest/dissect > index 16cc118..ce90bda 100755 > --- a/winetest/dissect > +++ b/winetest/dissect > @@ -252,9 +252,9 @@ while (<IN>) { > } elsif (/^(.*$unit.*: (\d+) tests executed $(\d+) marked as todo, (\d+) failures?$, (\d+) skipped\.)\r?$/) { > $lines++; > $total += $2; > - $todo += $3; > - $failed += $4; > - $skipped += $5; > + $todo = $3; > + $failed = $4; > + $skipped = $5; > chomp; > s/\r+$//; > my $class = "test result"; Hi André, That log also shows "server: 1 failures in child process". I'm not a 100% sure but from the looks of it this is that one/1 failure mentioned in that last line: server: 120 tests executed (1 marked as todo, 1 failure), 0 skipped. If I'm correct it does mean that we have 2 failures and 1 todo. (cc-ing AJ as he wrote that winetest_wait_child_process() thingy). -- Cheers, Paul.

3 6

Re: dmloader: Fix a possible null dereference
by Paul Vriens 05 Jan '10

05 Jan '10

On 01/05/2010 02:18 PM, David Adam wrote: > Hello > > This patch fixes the bug 13986 > > A+ > > David > Searc > Hi David, + hr = CoInitialize(NULL); + if ( FAILED(hr) ) + { + skip("CoInitializeEx failed.\n"); + return; + } The skip() message doesn't match the actual function called. + hr = CoCreateInstance(&CLSID_DirectMusicLoader, NULL, CLSCTX_INPROC, &IID_IDirectMusicLoader8, (void**)&loader); + if (FAILED(hr)) + { + skip("CoCreateInstance failed.\n"); + return; + } Your mixing tabs and spaces here. Shouldn't 'loader' be released on return? -- Cheers, Paul.

1 0

Re: Revert "dsound: Remove refcounting in classfactory."
by Alexandre Julliard 05 Jan '10

05 Jan '10

Maarten Lankhorst <m.b.lankhorst(a)gmail.com> writes: > This reverts commit 1b94450f789791908a370f16e125087b84f35bcb. > > See bug 21196 Again, when sending a revert you need to include a justification for why it must be reverted. Referencing a bug is not good enough, the git log must be understandable on its own. Also I don't see anything in that bug related to ref counting. -- Alexandre Julliard julliard(a)winehq.org

2 1

Version test crashing
by Stefan Dösinger 05 Jan '10

05 Jan '10

Hi, Since a few days my nightly cxtest runs show crashes in the version:info test: http://test.winehq.org/data/f74e312bf81032c46bd2ce080a5f6a33c7ac3ee3/wine_s… (Ubuntu 8.10, 64 bit build) Test.winehq.org has more results, for this wine build I am seeing 3 more timeouts(most likely because the crash error message wasn't ack'ed, so the process didn't exit). I don't know when this started, but I think it was somewhen mid-december. If needed I can run a regression test. The test runs in a clean wine prefix - my modified cxtest script cleans everything up, runs winetest, cleans up again, then runs a bunch of games and writes down the performance. Cheers, Stefan

2 2

Re: (try2)[PATCH 1/2] d3d10: Add a function to initialize the null variables.
by Alexandre Julliard 05 Jan '10

05 Jan '10

Rico Schüller <kgbricola(a)web.de> writes: > +static void d3d10_null_effect_variable_init(struct d3d10_effect_variable *v, struct ID3D10EffectVariableVtbl *vtbl) > +{ > + v->vtbl = vtbl; > + v->buffer = &null_local_buffer; > + v->type = &null_type; > +} > + > +void d3d10_null_effect_variable_init_all() > +{ > + d3d10_null_effect_variable_init(&null_local_buffer, (ID3D10EffectVariableVtbl *)&d3d10_effect_constant_buffer_vtbl); > + d3d10_null_effect_variable_init(&null_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_variable_vtbl); > + d3d10_null_effect_variable_init(&null_scalar_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_scalar_variable_vtbl); > + d3d10_null_effect_variable_init(&null_vector_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_vector_variable_vtbl); > + d3d10_null_effect_variable_init(&null_matrix_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_matrix_variable_vtbl); > + d3d10_null_effect_variable_init(&null_string_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_string_variable_vtbl); > + d3d10_null_effect_variable_init(&null_shader_resource_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_shader_resource_variable_vtbl); > + d3d10_null_effect_variable_init(&null_render_target_view_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_render_target_view_variable_vtbl); > + d3d10_null_effect_variable_init(&null_depth_stencil_view_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_depth_stencil_view_variable_vtbl); > + d3d10_null_effect_variable_init(&null_shader_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_shader_variable_vtbl); > + d3d10_null_effect_variable_init(&null_blend_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_blend_variable_vtbl); > + d3d10_null_effect_variable_init(&null_depth_stencil_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_depth_stencil_variable_vtbl); > + d3d10_null_effect_variable_init(&null_rasterizer_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_rasterizer_variable_vtbl); > + d3d10_null_effect_variable_init(&null_sampler_variable, (ID3D10EffectVariableVtbl *)&d3d10_effect_sampler_variable_vtbl); > +} > + That's not much of an improvement, static initialization is nicer. -- Alexandre Julliard julliard(a)winehq.org

2 2

Re: [2/5] WineD3D: Call the non-wrapped GLfinish in the PBO test
by Henri Verbeet 05 Jan '10

05 Jan '10

2010/1/4 Stefan Dösinger <stefan(a)codeweavers.com>: > In this case you can just release the GL lock, although the call is just an OS X hack in the first place.

1 0

Re: shell32: Initialize the APPBARDATA struct to avoid use of uninitialized memory in SHAppBarMessage.
by Paul Vriens 05 Jan '10

05 Jan '10

On 01/05/2010 02:47 AM, James Hawkins wrote: > --- > dlls/shell32/tests/appbar.c | 3 +++ > 1 files changed, 3 insertions(+), 0 deletions(-) > Hi James, Both Vincent and I tried something along those lines but AJ's response was: "That's just hiding the problem, an app can pass the same structure." and "We need to define a platform-independent struct for marshalling that data across processes anyway, we can't blindly send the structure." Vincent also mentioned http://bugs.winehq.org/show_bug.cgi?id=20982 -- Cheers, Paul.

1 0

Tango Icon Questions
by Joel Holdsworth 04 Jan '10

04 Jan '10

Ok, We seem to have two questions that need to be resolved before the tango icons can be merged. First: The SVGs. A little background on how I make the icon set: The icon set has been assembled from different sources of artwork. Some my own, some from other people, so there's a certain amount of variation in the formats that I've received. In every case, I have had to do a certain amount of work to produce the 9 or 10 sub-images of an ico. Rather like fonts, in icon art, the size of a pixel is significant relative to the size of graphical elements. This effect is particularly pronounced at the smaller sizes: 16x16 and 22x22. At smaller sizes, manual intervention in the vector artwork becomes increasingly necessary. Graphical elements have to be tweaked to align to pixel boundaries or resized, rearranged, or removed altogether. Low colour depths pose further problems. In 8-bit icons we only have 1-bit transparency. Drop shadows must be removed, and the icon must be adjusted in a raster editor to remove jagginess in the boundaries. These are exported as PNGs, and in some cases I kept the GIMP xcf.bz2s. 4-bit icons are still harder. Here the icon has to be so radically adjusted, that the original SVG really only forms a template for a piece of pixel art. I tend to do this work with the wine glass hidden in the SVG rendering, and then overlay a pre-rendered wine glass. Again, these are expored as PNGs, and in most cases I kept the GIMP xcf.bz2s. In an ideal world, we would have one SVG source file, which would be versioned in Git, which we would render to different sizes and depths as part of the build process. I hope you can see how this would be impractical because of the amount of manual raster work that has been needed. The only build scripts that I have are scripts that automatically compile my PNGs into ICOs, but this step wouldn't really help the project much, because the PNG is no less a binary blob than the ICO. To further compound the issue, I've found 2 bugs in icotool that cause it to create ICOs that are formatted incorrectly, and cause Wine to misbehave. For my ICO scripts I wrote a simple tool called "fix-icon" which tweaks the files, to correct the 2 types of problem. I've tried to contact the author of icotool about this, but as yet haven't got a response. So in conclusion, I think the best thing would be for me to tidy my icons project folder, and submit that to git or wine-pub-ftp. I think we need to accept that some images are PNG, some SVG and some XCF.BZ2, depending on the original source, and I don't think we can do automated render steps as part of the build. The other question is about attribution. By now, ALL the icons have had some input from me in one way or another; they're all a derivative work. I'm planning to just submit the icons in the current round, as shown on the website: http://www.airwebreathe.org.uk/wine-icon/ - no bitmaps this round, just shell32.dll, user32.dll, and Programs icons. The sources of these are: * Tango Base - Public Domain * Scott Ritchie - oic_winlogo.ico - the Wine Glass * Jakub Steiner - oic_bang.ico - GPLd, but I have an e-mail of him agreeing to relicense * Ulisse Perusin - winemine - GPLd, but have e-mail agreeing a relicense. * Joel Holdsworth - my original work - LGPLd. What would be the best way for us to attribute these? I guess we need to do things properly with regard to licensing, but I think the authors are all quite easy-going. Any thoughts Thanks Joel

1 0

Re: Patch feedback requested for OleCreatePropertyFrame()
by Geoffrey Hausheer 04 Jan '10

04 Jan '10

On Mon, Jan 4, 2010 at 9:06 AM, Wolfram Sang wrote: >> task. There doesn't appear to be anything in this code that would be >> patent encumbered, but the code itself certainly falls under >> Hidenori's copyright which he has effectively revoked permission to >> use. > > Strictly speaking, this is not possible. Although IANAL, this code is LGPL, > so he can't revoke permission for this version of the code. He can ask to > not use it, and this wish can be granted. You still should be allowed to > pick it up. Next question is then, if a modified version will be picked up > (still respecting the author's wish)? Or how much modifications are needed? > Geez, such things are never trivial :( > Well, I put the LGPL header on the code. I thought I had that right since the code was posted to wine-patches Anyhow, the legal side isn't really relevant. The question is what Alexandre will accept as far as it goes. The current code is certainly copyright by Hienori, and since I couldn't contact him, I must assume he doesn't want any of it in Wine. Thus my assumption is that I need to rewrite the entire thing, if we want to obey his wishes (which I do). However, my hope was that I could use the knowledge I gained from his implementation to do it myself. As I'm already biased by that knowledge, if that won't be allowed, then I can't do the work at all.

2 1