July 2009 - Wine-devel - list.winehq.org

AcceptEx Proposal
by Mike Kaplinskiy 05 Jul '09

05 Jul '09

I have been thinking about the correct way to implement AcceptEx, and I was looking for comments about the approach. Note I know all of these cases will require tests, I'm just listing them for problems someone might see. Wineserver changes: - add an async* to the socket structure to track this socket's request on the listening socket's queue. (set = null on accept) - Can anything do anything wrong while trying to remove the async when the accepting socket is closed before the accept event is received? (removing another socket's async request) - When/how should we handle setting this (create a new request type (that queues & sets), create a new request type (that just sets) or somehow splice into sock_queue_async? - my vote for first, as much as I hate it) - change accept_socket to use two sockets instead of returning the accepted socket - Problem: we have to destroy the fd of the old socket. Can anything queue async reads/writes before the overlapped acceptex returns? Do we care that the fd of a socket object gets switched (you lose internal state, but I don't think you can have a lot of state on an unbound socket) - Deferred sockets are a pain: 2 clients attempt to connect->accept->defer->acceptex=which socket does it return? Also if this order happens, and we have to destroy the fd as above, would the application possibly care? - need to implement SO_UPDATE_ACCEPT_CONTEXT and have accept() call it right after getting an accept. - if we close the listener socket, what happens to sockets waiting for clients on the accepting socket? Can we reuse these sockets? ws2 changes: - exactly the same as the previous attempt, just using the above requests - more strict checks are needed on the second acceptex socket (need to check family, bound state) - accept: a = create_socket(), accept_socket(listener, a), ioctl)SO_UPDATE_ACCEPT_CONTEXT, a, &listener); I just started looking at wineserver, so don't be surprised if something above doesn't make sense. Mike.

2 2

Howto put a new rpm package on sourceforge ?
by marco 05 Jul '09

05 Jul '09

Hi all, I make the mandriva packages and put them on sourceforge. But sourceforge changed there layout and now I can not longer find the button to add new releases or to manage them? Anyone know where they are ? The sourgeforge wiki is taking about a admin button on the summary page, but it is not visible when I log in. Can someone help me ? kind regards, Marco Meijer

3 4

16-bit Code on MacOSX Leopard/XCode 3.1
by James McKenzie 05 Jul '09

05 Jul '09

I know that this was disabled, but what happens when I try to run a 16 bit app? Here is what happened when I tried to build 1.1.25 with XCode 3.1 today: checking whether 16-bit code can be built correctly... no configure: error: Xcode 3.x cannot build 16-bit code correctly. Use --disable-win16 if you don't need 16-bit support. ***** Error: Can't build wine Great catch, BTW. James McKenzie

3 3

Re: fix overflow in several advapi32 crypt functions hidden by broken WideCharToMultiByte
by James McKenzie 04 Jul '09

04 Jul '09

Christoph von Wittich wrote: As Dmitry wrote about this, I will add the comment that this does not look like the proper way to fix this problem. The proper way is to keep the pointer from overflowing in the first place and that would be to place a check where it is set and updated to either go negative (which should not be allowed) or to be incremented past INT_MAX. This can be implemented in several different ways, and since I have not looked at the code, I could not make a proper suggestion but would suggest that the pointer be checked before incrementing and after setting. James McKenzie

1 0

Re: Ensure that GetSystemMenu returns valid HMENU unless bRevert specified
by James McKenzie 04 Jul '09

04 Jul '09

Brent Roman wrote: > Originally submitted on 6/14/09. > Resubmitted after comments on this bug: > > http://bugs.winehq.org/show_bug.cgi?id=19134 > > The attached patch against the master branch (as of today) fixes what > appears to be a long standing bug in Wine's > GetSystemMenu function in User32.dll.so. > > GetSystemMenu() is always supposed to return a valid HMENU handle, but > the logic in Wine's version would return nil in some cases where > windows would return an HMENU. Specifically, when the relevant > windows dwStyle mask did not include WS_SYSMENU, wine would return > nil, but windows would still return a valid HMENU. > > To reproduce the bug: > > 1. Download and install the Quadravox AQ430 development tools from: > http://www.quadravox.com/AQ430.htm > http://appdb.winehq.org/objectManager.php?sClass=application&iId=9939 > > 2. Attempt to compile any 'C' source file. > > Before applying the patch, all such attempts will quickly fail with an > invalid read from address 0x4. You rewrote the entire function, there is no need to do this. Please use the same formatting as the rest of the code. Add new variables to the bottom of the list, you imbedded one or more. Since the rest of the function uses four spaces, use them. Avoid the use of tabs, some editors spread them to four spaces, some two and yet others eight. Makes the code look sloppy and is very hard to read. Can you provide test cases, other than 'use this program to prove it works'? The idea is to replicate Windows API functionality and some programs rely on certain functionality working in a very specific manner. If you build to the program, you might break others. Also, if you have left out some function, please insert c style notes stating this. Makes it easier to find where certain code pieces go at a later time. James McKenzie

1 0

A tiny GetSystemMetrics question
by Nikolay Sivov 04 Jul '09

04 Jul '09

Hi. Trying to fix some ListView bugs I've found that on native (XP SP2) system this call: --- GetSystemMetrics(SM_CXICONSPACING) --- doesn't return the value specified in appearance -> advanced settings of Display properties sheet. I've got 75 value for both cx and cy while I have set it to 43 in dialog box. This look weird for me. Does anybody know what is the reason of this behavior? P.S. Yes, I'm using default 96 dpi on this XP box (if it does matter).

2 2

Re: [PATCH 3/5] d3d9/tests: Add a small test for having multiple device active at the same time.
by Paul Vriens 03 Jul '09

03 Jul '09

Henri Verbeet wrote: > This is essentially the situation that caused problems with reusing the > initial GL context. > --- > dlls/d3d9/tests/device.c | 65 ++++++++++++++++++++++++++++++++++++++++++++++ > 1 files changed, 65 insertions(+), 0 deletions(-) > Hi Henri, > + hr = IDirect3D9_CreateDevice(d3d9, D3DADAPTER_DEFAULT, D3DDEVTYPE_HAL, hwnd1, > + D3DCREATE_SOFTWARE_VERTEXPROCESSING, &present_parameters, &device1); > + ok(SUCCEEDED(hr), "Failed to create a device, hr %#x\n", hr); This one seems to fail on a lot of boxes, currently only VMware and ATI so it seems. NVIDIA and Intel seem to be fine though (for now as only a few tests are in). Is D3DERR_NOTAVAILABLE an acceptable error? If so than the ok() message can be easily extended (maybe with an extra skip() message?). Could you have a look? -- Cheers, Paul.

2 1

user32: Windows test request (cursors/icons)
by Daniel Santos 03 Jul '09

03 Jul '09

I don't have access to a working windows machine right now and would appreciate if somebody can run these tests against XP, vista and/or 9x. Thanks! Daniel

2 1

Re: [2/2] gdiplus: Implement GdipBeginContainer2 and GdipEndContainer
by Nikolay Sivov 03 Jul '09

03 Jul '09

Andrew Eikum wrote: > --- > dlls/gdiplus/gdiplus_private.h | 3 + > dlls/gdiplus/graphics.c | 154 ++++++++++++++++++++++++++++++++++++++-- > include/gdiplusflat.h | 1 + > 3 files changed, 152 insertions(+), 6 deletions(-) > This looks wrong for me cause I don't think you could pass container id created with one Graphics to Graphics.EndContainer() of another. I've just checked it and it doesn't return any error code, but generated container ids aren't the same when you call BeginContainer() for e.g. two objects as sequent steps. It looks like on native these value are module unique, some kind of handle table needed. This kills the first problem I described - passing strange value to another Graphics. Also note that Save()/Restore() functionality uses the same stack with container calls (at least msdn tells us that), for that Save/Restore we already have a test: --- static void test_save_restore(void) .... /* The same state value should never be returned twice. */ todo_wine check_no_duplicates(state_log); .... --- It clearly says that no duplicates allowed, even after multiple object deletion - it checks for values returned during the whole test.

3 7

Re: kernel32: Compile .mc files to resources as independent files.
by Paul Vriens 03 Jul '09

03 Jul '09

Hi, kernel32 translations don't show up at the translation page anymore. Do we need to teach wrc to deal with rc files being one level lower (nls directory?) or what is needed to correct this? (Before this patch the page basically showed if there was a translation or not and not whether the content was correct) -- Cheers, Paul.

2 1