wine-bugs November 2007

wine-bugs@winehq.org

19 participants
2259 discussions

[Bug 5120] menu closes on mouseover
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=5120 Vitaliy Margolen <vitaliy(a)kievinfo.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution| |DUPLICATE --- Comment #4 from Vitaliy Margolen <vitaliy(a)kievinfo.com> 2007-11-26 23:55:32 --- Duplicate *** This bug has been marked as a duplicate of bug 758 *** -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 6767] failure enumerating interfaces on NetBSD
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=6767 jmmikkel(a)mit.edu changed: What |Removed |Added ---------------------------------------------------------------------------- Attachment #9361|standalone enumIP |standalone enumIPAddresses description| |tester --- Comment #7 from jmmikkel(a)mit.edu 2007-11-26 23:48:12 --- (From update of attachment 9361) this is the code I used to test proposed changes -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 6767] failure enumerating interfaces on NetBSD
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=6767 --- Comment #6 from jmmikkel(a)mit.edu 2007-11-26 23:47:02 --- Hi, I apologize for having dropped the ball back when. Then, this past June I set out to deal with this and found that the code had changed. Unfortunately the current code will still not work correctly, and the application I was using when this problem came up for me no longer worked but for a different reason (it does not even get this far). I didn't have another application that used this functionality to try with, and I have not managed to sort through it since then. (I do use the pkgsrc build, although it would appear 0.9.39 compiled without pkgsrc patches in June.) The problem with the current version is that NetBSD places the required length in ifc->ifc_len at all times. So it will always have the same value two ioctl calls in a row, even if the provided buffer is still too small the second time. Because I don't have an easy way to test this with an actual application in Wine, I have copied the enumIPAddresses function to a separate file, and added some printing, so I can test with just that. When I run the existing code, I find that ifc->ifc_len starts at 128, and the ioctl sets it to 352. Since ifc->ifc_len != lastlen (352 != 0) it goes through the loop again, setting lastlen to 352 and ifc->ifc_len to 256. The ioctl once again sets it to 352 and the loop terminates, but it shouldn't have, because 352 > 256 so my list will be truncated. I don't know why it was changed to do the "two in a row" test. I can't think of a way for the ioctl to work that would make that test necessary instead of something more like the >= test I originally suggested. The ioctl can put three kinds of numbers in ifc_len. Numbers smaller than the provided buffer indicate the total length of the data, including zero. This means that for cases where the buffer is not big enough, there are only two possible choices: 1.) ioctl() returns the buffer size if the buffer is too short (Linux behavior, IIRC) 2.) ioctl() returns a number bigger than the size if the buffer is too short (NetBSD -- we actually need not assume the number is accurate). The twice-in-a-row solution we currently have always results in an extra trip through the loop unless the actual length of the interface list is zero, because lastlen starts out at zero. Consider a required length of 100. lastlen is set to 0, the buffer is allocated with size 128, and the ioctl sets the required length to 100. This is != 0 so we loop again, setting lastlen to 100 and allocating a buffer of size 256. The ioctl sets the length to 100 again and the loop terminates, having run twice instead of once. At any rate, the solution certainly covers situation 1: 128 is < 352 so the ioctl sets the length to 128, and the loop sets lastlen to 128. The ioctl is called with a buffer of 256 and ioctl sets the length to 256, which is != 128, so we loop again and call with a buffer of 512. The ioctl sets the length to 352, which is != 256, and we call the ioctl again with a buffer of 1024 but then terminate. The twice-in-a-row solution doesn't cover situation 2 if the required length is > 256 as I described above. It works okay when the length is less than 256. Perhaps this is the situation that was addressed with the change to the current algorithm? Oddly, the solution also works if the ioctl does something strange like fill in the provided size + 1 when the buffer is not big enough (buffer size 128 -> ifc_len set to 129, buffer size 256 -> ifc_len set to 257, buffer size 512 -> ifc_len set to 352, buffer size 1024 -> ifc_len set to 352). But it breaks if the ioctl is *consistent* about the size. A >= test also covers situation 1, at the price of an extra trip through the loop but only when the required buffer size is exactly what was provided. But that corner case required an extra trip in the first version I worked with too (using the == test you can see in the diff in the initial report). If 352 is required and 128 is provided, the ioctl sets the length to 128 and 128 >= 128 so the next loop the buffer size is 256 and the ioctl sets it to 256 and 256 >= 256 so the next loop the buffer size is 512 and the ifc_len is set to 352, which is not >= 512 and the loop terminates. The >= test covers situation 2 as well, including for required buffer lengths > 256, of course, since it's the fix that worked for me. For NetBSD we would call the ioctl with a buffer size of 128, the ifc_len would be set to 352, and 352 >= 128 so the next time through the loop would call ioctl with a size of 256, ifc_len would be set to 352, which is >= 256, so the next loop the buffer size is 512 and the ioctl sets ifc_len to 352, which is not >= 512. This also works with the strange version of the ioctl that returns the size + 1 when the provided size is too small: buffer size 128 -> ifc_len set to 129 and 129 >= 128 so now buffer size 256 -> ifc_len 257 and 257 >= 256 so then buffer size 512 -> ifc_len set to 352, which is not >= 512 and the loop terminates. So my suggestion would be to return the loop back to how it was a year ago but with >= instead of == (removing lastlen): ret = NO_ERROR; ifc->ifc_len = 0; ifc->ifc_buf = NULL; /* there is no way to know the interface count beforehand, so we need to loop again and again upping our max each time until returned < max */ do { HeapFree(GetProcessHeap(), 0, ifc->ifc_buf); if (guessedNumAddresses == 0) guessedNumAddresses = INITIAL_INTERFACES_ASSUMED; else guessedNumAddresses *= 2; ifc->ifc_len = sizeof(struct ifreq) * guessedNumAddresses; ifc->ifc_buf = HeapAlloc(GetProcessHeap(), 0, ifc->ifc_len); ioctlRet = ioctl(fd, SIOCGIFCONF, ifc); } while (ioctlRet == 0 && ifc->ifc_len >= (sizeof(struct ifreq) * guessedNumAddresses)); But if the twice-in-a-row test is really necessary for some reason I'm not seeing here, we could perhaps patch up NetBSD with the following diff to the current version, which will take advantage of the provided required length without changing the algorithm for OSes that leave the provided buffer size as the length when it's long enough. It seems more hackish to me, though (but it does eliminate the extra trip through the loop when the required size is >= 128!). --- dlls/iphlpapi/ifenum.c.orig +++ dlls/iphlpapi/ifenum.c @@ -695,6 +695,9 @@ else guessedNumAddresses *= 2; ifc->ifc_len = sizeof(struct ifreq) * guessedNumAddresses; + if (lastlen > ifc->ifc_len) { + ifc->ifc_len = lastlen; + } ifc->ifc_buf = HeapAlloc(GetProcessHeap(), 0, ifc->ifc_len); ioctlRet = ioctl(fd, SIOCGIFCONF, ifc); } while ((ioctlRet == 0) && (ifc->ifc_len != lastlen)); Sorry I can't test this with the application I was using any more (I'd have to debug some other thing that has gone wrong in the last year first) but I can easily test out whatever change is decided upon with my little standalone tester. -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 6767] failure enumerating interfaces on NetBSD
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=6767 --- Comment #5 from jmmikkel(a)mit.edu 2007-11-26 23:44:34 --- Created an attachment (id=9361) --> (http://bugs.winehq.org/attachment.cgi?id=9361) standalone enumIP -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 9036] Adobe Photoshop Elements 1.0 fails to install, says IE4 is required
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=9036 Fred P. <fpsys2004-wine(a)yahoo.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |fpsys2004-wine(a)yahoo.com -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 7831] Snood: not displayed correctly and program freezes after first shot
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=7831 Nick Schubach <nschubach(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |nschubach(a)gmail.com --- Comment #3 from Nick Schubach <nschubach(a)gmail.com> 2007-11-26 20:19:33 --- I ran across the same issue. I downloaded the latest "shareware" version and was able to shoot one time, then it seemed like the mouse was non-responsive. I couldn't click the menu items or anything after the first shot. If I run it from the command line, I don't even get one shot, it loses the mouse (and keyboard) after clicking the "New Game" button and I get only one error: fixme:win:EnumDisplayDevicesW ((null),0,0x34f870,0x00000000), stub! It wouldn't be a big issue except the alternatives (Frozen-Bubble, xbobble) don't seem to support the mouse usage, and it makes controlling MUCH nicer with the mouse. -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 3542] Dungeon Keeper Gold not starting
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=3542 Stewart Adam <s.adam(a)diffingo.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |s.adam(a)diffingo.com --- Comment #18 from Stewart Adam <s.adam(a)diffingo.com> 2007-11-26 19:41:53 --- I am seeing the same issue, nvidia driver 100.14.19 on Fedora 8, wine 0.9.49. Running the full path (wine "C:\Program Files\DKGold\KEEPER95.EXE") allows it to run properly, but it hangs right after the intro movie (X locks up so I could not obtain any errors). -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 6250] Mouse cursor is not visible
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=6250 Jared Sutton <jpsutton(a)gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jpsutton(a)gmail.com -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 3591] py2exe does not work properly
by wine-bugs＠winehq.org 27 Nov '07

27 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=3591 Nerijus Baliunas <nerijus(a)users.sourceforge.net> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |nerijus(a)users.sourceforge.ne | |t Version|20050930 |0.9.49. --- Comment #5 from Nerijus Baliunas <nerijus(a)users.sourceforge.net> 2007-11-26 18:19:35 --- BeginUpdateResource & EndUpdateResource are now implemented (bug 3897), but one more is missing - BindImageEx. When using py2exe I get: *** finding dlls needed *** fixme:imagehlp:BindImageEx (7, "C:\\python25\\DLLs\\_hashlib.pyd", ... C:\\python25\\l"..., (null), 0x541600): stub ... *** create binaries *** *** byte compile python files *** ... and it finishes succesfully. But the strange thing is that I can successfully run generated exe file under wine and Win98, but under XP trying to run it gives: app.exe is not a valid Win32 application. -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

[Bug 6921] C&C Tiberian Sun menu disappears after save
by wine-bugs＠winehq.org 26 Nov '07

26 Nov '07

http://bugs.winehq.org/show_bug.cgi?id=6921 --- Comment #6 from Alexander Nicolaysen Sørnes <alex(a)thehandofagony.com> 2007-11-26 16:31:43 --- The game also stops re-drawing if an Ion Cannon is fired when the game is run in opengl mode. Red Alert 2 also stops re-drawing when certain actions are performed, like when the stats screen is displayed after a skirmish. -- Configure bugmail: http://bugs.winehq.org/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are watching all bug changes.

1 0

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

wine-bugs November 2007