November 2009 - Wine-Devel - list.winehq.org

Re: Wine on OpenBSD, Theo de Raadt

by Mihai Donțu

On Thursday 05 November 2009 19:44:47 David Gerard wrote: > 2009/11/5 Mihai Donțu <mihai.dontu(a)gmail.com>: > > I've never been able to run a dos program in Wine, but I've never asked > > myself why because I always had an alternative. > > DOSbox is a much better DOS than Wine is ever likely to be. > > Do Win32 console apps run okay in Wine without the page 0 mapping? I > I use wine to run the Microsoft Visual Studio Express compiler (cl.exe). So far I hadn't had any problems. > ask because in my last job, we actually used Wine to run a Win32 > console app on RHEL4 to save installing another Windows box. Never let > anyone tell you Wine isn't ready for the enterprise ;-) > -- Mihai Donțu

Nov. 5, 2009

Re: Wine on OpenBSD, Theo de Raadt

by David Gerard

2009/11/5 Mihai Donțu <mihai.dontu(a)gmail.com>: > I've never been able to run a dos program in Wine, but I've never asked myself > why because I always had an alternative. DOSbox is a much better DOS than Wine is ever likely to be. Do Win32 console apps run okay in Wine without the page 0 mapping? I ask because in my last job, we actually used Wine to run a Win32 console app on RHEL4 to save installing another Windows box. Never let anyone tell you Wine isn't ready for the enterprise ;-) - d.

Nov. 5, 2009

Re: Wine on OpenBSD, Theo de Raadt

by Mihai Donțu

On Thursday 05 November 2009 18:31:52 Yann Droneaud wrote: > Le jeudi 05 novembre 2009 à 17:01 +0100, Marcus Meissner a écrit : > > On Thu, Nov 05, 2009 at 02:21:47PM +0100, Yann Droneaud wrote: > > > Hi, > > > > > > In a recent post ( <http://lwn.net/Articles/360312/> ), Theo de Raadt > > > is criticising Linux page 0 handling. > > > He argue that Linux allows page 0 mapping only for Wine. > > > And OpenBSD won't allow page 0 mapping for security reasons (and > > > performance reasons). > > > > > > Since Wine seems to be working on other *BSD and derivative systems, > > > is this a problem ? > > > > > > Could someone comment on this issue. I don't find anything related on > > > <http://wiki.winehq.org/OpenBSD> > > > > Its actually not a real problem for Wine, last years code works without > > NULL pages. > > It seems many people thinks Wine needs to mmap the page 0 to run common > application. > > Does Wine 1.0 need it ? > > If not, it's really a myth. > # sysctl vm.mmap_min_addr vm.mmap_min_addr = 4096 and wine works like a charm. > > Our DOS emulator requires it, as vm86() mode uses 0 based mapping, > > so DOS programs won't work afterwards. But otherwise I see no dependency. > > (Perhaps Win16, but I dont think so.) > I've never been able to run a dos program in Wine, but I've never asked myself why because I always had an alternative. > According to this post from Alexandre Julliard, win16 don't need it. > > http://old.nabble.com/Re%3A-vm86-mode-is-not-supported-p23884662.html > > Perhaps it could be good point if there's a way to disable DOS emulator > at build time ? > -- Mihai Donțu

Nov. 5, 2009

Re: Wine on OpenBSD, Theo de Raadt

by Marcus Meissner

On Thu, Nov 05, 2009 at 05:31:52PM +0100, Yann Droneaud wrote: > Le jeudi 05 novembre 2009 à 17:01 +0100, Marcus Meissner a écrit : > > On Thu, Nov 05, 2009 at 02:21:47PM +0100, Yann Droneaud wrote: > > > Hi, > > > > > > In a recent post ( <http://lwn.net/Articles/360312/> ), Theo de Raadt is > > > criticising Linux page 0 handling. > > > He argue that Linux allows page 0 mapping only for Wine. > > > And OpenBSD won't allow page 0 mapping for security reasons (and > > > performance reasons). > > > > > > Since Wine seems to be working on other *BSD and derivative systems, > > > is this a problem ? > > > > > > Could someone comment on this issue. I don't find anything related on > > > <http://wiki.winehq.org/OpenBSD> > > > > Its actually not a real problem for Wine, last years code works without > > NULL pages. > > > > It seems many people thinks Wine needs to mmap the page 0 to run common > application. > > Does Wine 1.0 need it ? > > If not, it's really a myth. I installed 1.0.1 , enabled mmap_min_addr protectionm ... Testcase: Firefox 3.5.4 ... installs and runs nicely. (OK, I had some crashes after Flash Player installation for youtube, but youtube plays). > According to this post from Alexandre Julliard, win16 don't need it. > > http://old.nabble.com/Re%3A-vm86-mode-is-not-supported-p23884662.html > > Perhaps it could be good point if there's a way to disable DOS emulator o> at build time ? You do not know that at build time. My WINE rpms build for i586 can be run on i586 and x86_64 machines. Ciao, Marcus

Nov. 5, 2009

Re: Wine on OpenBSD, Theo de Raadt

by Yann Droneaud

Le jeudi 05 novembre 2009 à 17:01 +0100, Marcus Meissner a écrit : > On Thu, Nov 05, 2009 at 02:21:47PM +0100, Yann Droneaud wrote: > > Hi, > > > > In a recent post ( <http://lwn.net/Articles/360312/> ), Theo de Raadt is > > criticising Linux page 0 handling. > > He argue that Linux allows page 0 mapping only for Wine. > > And OpenBSD won't allow page 0 mapping for security reasons (and > > performance reasons). > > > > Since Wine seems to be working on other *BSD and derivative systems, > > is this a problem ? > > > > Could someone comment on this issue. I don't find anything related on > > <http://wiki.winehq.org/OpenBSD> > > Its actually not a real problem for Wine, last years code works without > NULL pages. > It seems many people thinks Wine needs to mmap the page 0 to run common application. Does Wine 1.0 need it ? If not, it's really a myth. > > Our DOS emulator requires it, as vm86() mode uses 0 based mapping, > so DOS programs won't work afterwards. But otherwise I see no dependency. > (Perhaps Win16, but I dont think so.) > According to this post from Alexandre Julliard, win16 don't need it. http://old.nabble.com/Re%3A-vm86-mode-is-not-supported-p23884662.html Perhaps it could be good point if there's a way to disable DOS emulator at build time ? Regards -- Yann Droneaud

Nov. 5, 2009

Re: crypt32.dll: Initialize provider pointer to null before calling CryptAcquireContextW()

by Juan Lang

> Checking the return value of CryptAcquireContextW() is also possible but I > took the path of minimal change in the source. The change I suggest would be small too. > Besides, I consider passing pointers to uninitialized variables a bad > programming practice. Not when they're out pointers. --Juan

Nov. 5, 2009

Re: Wine on OpenBSD, Theo de Raadt

by Marcus Meissner

On Thu, Nov 05, 2009 at 02:21:47PM +0100, Yann Droneaud wrote: > Hi, > > In a recent post ( <http://lwn.net/Articles/360312/> ), Theo de Raadt is > criticising Linux page 0 handling. > He argue that Linux allows page 0 mapping only for Wine. > And OpenBSD won't allow page 0 mapping for security reasons (and > performance reasons). > > Since Wine seems to be working on other *BSD and derivative systems, > is this a problem ? > > Could someone comment on this issue. I don't find anything related on > <http://wiki.winehq.org/OpenBSD> Its actually not a real problem for Wine, last years code works without NULL pages. Our DOS emulator requires it, as vm86() mode uses 0 based mapping, so DOS programs won't work afterwards. But otherwise I see no dependency. (Perhaps Win16, but I dont think so.) Ciao, Marcus

Nov. 5, 2009

Wine on OpenBSD, Theo de Raadt

by Yann Droneaud

Hi, In a recent post ( <http://lwn.net/Articles/360312/> ), Theo de Raadt is criticising Linux page 0 handling. He argue that Linux allows page 0 mapping only for Wine. And OpenBSD won't allow page 0 mapping for security reasons (and performance reasons). Since Wine seems to be working on other *BSD and derivative systems, is this a problem ? Could someone comment on this issue. I don't find anything related on <http://wiki.winehq.org/OpenBSD> Regards. -- Yann Droneaud

Nov. 5, 2009

Re: crypt32.dll: Initialize provider pointer to null before calling CryptAcquireContextW()

by Paul Chitescu

On Wednesday 04 November 2009 11:45:38 pm Juan Lang wrote: > Hi Paul, > > I imagine this is to avoid checking prov if CryptAcquireContextW > fails. In that case, wouldn't it be better to check the return value > of CryptAcquireContextW? > --Juan Checking the return value of CryptAcquireContextW() is also possible but I took the path of minimal change in the source. Besides, I consider passing pointers to uninitialized variables a bad programming practice. Paul

Nov. 5, 2009

Re: Weird error for the advapi32/eventlog tests on Vista+ when running winetest

by Paul Vriens

On 11/05/2009 10:11 AM, Paul Vriens wrote: > On 11/04/2009 09:45 PM, Reece Dunn wrote: >> 2009/11/4 Paul Vriens<paul.vriens.wine(a)gmail.com>: >>> Hi, >>> >>> Just noticed that we have 1 test failure on Vista+ for the >>> advapi32/eventlog >>> tests. I was sure I've tested these from Win95 up to Win7. >>> >>> When I run tests I want to do this as Administrator, so on Vista+ I >>> select >>> 'Run as Administrator' to start a command prompt, or to launch winetest. >>> >>> Running the mentioned crosscompiled tests is no problem unless you do >>> this >>> from the actual home directory (or any subdirectory) of the user that >>> logged >>> in. >>> >>> Winetest runs most of it's test in the %TEMP% directory and hence we see >>> this failure. But I have no clue why (yet). >>> >>> Anyone? >> > > Hi Reece, > >> Ah, I see. This is the failure on test.winehq.org, correct? > > Yes. I'm in the process of writing loads of eventlog tests while keeping > track of possible failures on test.winehq.org. > >> >> eventlog.c:586: Test failed: Expected ERROR_INVALID_HANDLE, got 5 >> >> 5 is ERROR_ACCESS_DENIED. >> >> The create_backup function does not have any error checking -- what >> happens when you put the return values for these calls, and the >> GetLastError in ok checks, e.g. >> >> ok(DeleteFileA(filename), "error deleting the backup event log file"); >> ok(GetLastError() == 0, "GetLastError -- expected ERROR_SUCCESS, >> got %d", GetLastError()); >> >> - Reece > > Deleting the file works just fine. both in create_backup() and > test_clear(). > > What makes it even more weird is that I have the same error when I run > these tests as the user (not admin). > > I also don't get any UAC prompts in either case. Turning UAC off doesn't > make a difference. > > I don't see any logical explanation why the function returns a different > error. The fact that the function fails anyway makes me consider using a > second possible last error for that particular test. > Found it. The eventlog service runs under an account that doesn't have the necessary rights for a users home directory. This account is the actual one performing some of the actions. -- Cheers, Paul.

Nov. 5, 2009