July 2008 - Wine-devel - list.winehq.org

valgrind progress: down to 164 errors in 52 tests
by Dan Kegel 18 Jul '08

18 Jul '08

This morning a whole bunch of warnings in directx, dde, gdiplus, and winmm were fixed; see http://kegel.com/wine/valgrind/logs/2008-07-18-07.39-summary.txt This brings us down from yesterday's 187 errors in 67 tests to 164 errors in 52 tests. That's a nice way to start the day.

1 0

Re: Re[2]: advapi32: "RegGetValue" fails with some correct combinations of "dwFlags"
by James Hawkins 18 Jul '08

18 Jul '08

2008/7/18 Mathias Kosch <info(a)mkosch.de>: > Alexandre Julliard wrote: > >> Patches should be generated from the top-level directory. > Sorry for that, I fixed it. Please read [1] to find out how to properly send patches in Wine. You need to send one patch per email. [1] http://www.winehq.org/site/sending_patches -- James Hawkins

1 0

re: ntdll: Remove byte reversed U+3000 (CJK space) from being checked in RtlIsTextUnicode.
by Dan Kegel 18 Jul '08

18 Jul '08

Hang on, if you don't check U+3000 in the big endian case, you probably shouldn't check it in the little endian case, either, right?

2 1

Re: fix async read on mailslots
by Rob Shearman 18 Jul '08

18 Jul '08

2008/7/18 Ambroz Bizjak <ambro(a)b4ever.net>: > Hi, > In Wine creating a mailslot (CreateMailslot) with non-zero timeout > (argument 3 to CreateMailslot) and then attempting an asynchronous read on > it will cause a synchronous read to be performed - ReadFile will block > instead of failing with ERROR_IO_PENDING. Good spot, but then the correct fix is to only set FILE_SYNCHRONOUS_IO_NONALERT on the created fd if the timeout is non-zero. Asynchronous I/O has a performance penalty in Wine (extra server round-trips) so we want to avoid it if at all possible. > I'm attaching a fix as suggested by jil and a test. > The test creates an empty mailbox with a small timeout and performs an > asynchronous read - if broken, it will fail with ERROR_SEM_TIMEOUT. > I've also tried the test on Vista. It's good that you've included a test, but a few comments: > + mailslot = CreateMailslot("\\\\.\\mailslot\\TestMailslot", 100, 10, NULL); > + ok(mailslot!=INVALID_HANDLE_VALUE, "CreateMailslot failed\n"); > + if(mailslot==INVALID_HANDLE_VALUE) return; > + > + event = CreateEvent(NULL, FALSE, FALSE, NULL); > + ok(event!=INVALID_HANDLE_VALUE, "CreateEvent failed\n"); > + if(event==INVALID_HANDLE_VALUE) return; > + > + ol.Offset = 0; > + ol.OffsetHigh = 0; > + ol.hEvent = event; > + res = ReadFile(mailslot, buf, 100, NULL, &ol); > + ok(res==FALSE, "ReadFile succeeded; expected failure\n"); > + if(res!=FALSE) return; > + > + res = GetLastError(); > + ok(res==ERROR_IO_PENDING, "ReadFile returned error code %d, expected ERROR_IO_PENDING\n", res); There's no need to return after every unexpected failure. If one of the tests is likely to fail on some machine due to a configuration issue or some other feature beyond our control then it is acceptable to put in a skip statement, but generally we only do this after we've seen real-world evidence of failures. You also didn't include an LGPL header for the new file that you added. Also, in general you should include only one patch per mail to wine-patches and include an appropriate changelog (I see that you are using git and there are various tools to do this automatically). It is helpful if you can submit a test that has appropriate todo_wine's that show where Wine is failing a test, and then to submit a patch which does the fix and removes the todo_wine. However, in this case I don't believe that it is possible so you can send the fix first and then the tests. -- Rob Shearman

1 0

Re: [2/6] winhttp: Allow winhttp to use wininet.h with winhttp.h
by James Hawkins 18 Jul '08

18 Jul '08

On Thu, Jul 17, 2008 at 7:37 PM, Zac Brown <zac(a)zacbrown.org> wrote: > Setup winhttp.h and wininet.h with type guards to allow for use of wininet.h > in winhttp implementation. > > --snip-- > > Note: I am unsure of whether its better to do this, or finish up winhttp.h > and then create a private header in dlls/winhttp for winhttp to use. > diff --git a/dlls/winhttp/Makefile.in b/dlls/winhttp/Makefile.in index e69b137..7e8146f 100644 --- a/dlls/winhttp/Makefile.in +++ b/dlls/winhttp/Makefile.in @@ -4,7 +4,7 @@ SRCDIR = @srcdir@ VPATH = @srcdir@ MODULE = winhttp.dll IMPORTLIB = winhttp -IMPORTS = kernel32 +IMPORTS = wininet kernel32 C_SRCS = main.c I thought the idea was to implement winhttp and then implement wininet on top of winhttp. Why are you importing wininet? -- James Hawkins

2 1

Re: advapi32: "RegGetValue" fails with some correct combinations of "dwFlags"
by Alexandre Julliard 18 Jul '08

18 Jul '08

Mathias Kosch <info(a)mkosch.de> writes: > --- old/registry.c 2008-07-11 17:55:55.000000000 +0200 > +++ registry.c 2008-07-18 01:39:21.000000000 +0200 > @@ -1466,6 +1466,8 @@ Patches should be generated from the top-level directory. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: kernel32 [1/2]: Don't try to convert a possibly uninitialized out param on error
by Alexandre Julliard 18 Jul '08

18 Jul '08

"James Hawkins" <jhawkins(a)codeweavers.com> writes: > Changelog: > * Don't try to convert a possibly uninitialized out param on error. There are several ways to fix that one, it will need test cases to find out which one is right. -- Alexandre Julliard julliard(a)winehq.org

1 0

Re: [5/6] winhttp: main.c: Add implementation for WinHttpOpenRequest
by Dan Hipschman 17 Jul '08

17 Jul '08

On Thu, Jul 17, 2008 at 05:38:29PM -0700, Zac Brown wrote: > > + if (dwFlags & WINHTTP_FLAG_ESCAPE_PERCENT || > + dwFlags & WINHTTP_FLAG_NULL_CODEPAGE || > + dwFlags & WINHTTP_FLAG_ESCAPE_DISABLE || > + dwFlags & WINHTTP_FLAG_ESCAPE_DISABLE_QUERY) if (dwFlags & (WINHTTP_FLAG_ESCAPE_PERCENT | WINHTTP_FLAG_NULL_CODEPAGE | WINHTTP_FLAG_ESCAPE_DISABLE | WINHTTP_FLAG_ESCAPE_DISABLE_QUERY)) is shorter and more efficient since it boils down to one "test" instruction (at least it should), rather than a bunch of branching required by short-circuiting.

1 0

Re: ws2_32: update to WSASendTo / last error patch
by James Hawkins 17 Jul '08

17 Jul '08

On Thu, Jul 17, 2008 at 2:19 PM, Daniel Santos <daniel(a)bytesage.com> wrote: > Since this is my 1st real patch for wine, it's taken me a few iterations to > get this down. Attached is the updated patch to WSASendTo that includes a > conformance test. I have NOT run this on windows yet because I don't have > access to a windows machine at the moment. > Then we can't accept the tests (and thus the implementation). In the future, please attach your test patch to a message in wine-devel asking someone to run the test in windows. -- James Hawkins

2 1

Re: advapi32: "RegGetValue" fails with some correct combinations of "dwFlags"
by Mathias Kosch 17 Jul '08

17 Jul '08

Dan Kegel wrote: > You forgot to mention which versions of Windows you > ran the test case on (and whether it passed there). Thank you for telling me. (I'm not yet very familar with the process of publishing patches.) It took me some time, for I first had to install "Windows Server 2003". Testing was a good idea because I found out that my bugfix was a little too lax. Windows only allows "RRF_RT_ANY" without "RRF_NOEXPAND", but not a combination of "(RRF_RT_REG_EXPAND_SZ | RRF_RT_REG_SZ)". This is still a little curious because I thought "dwFlags" would act like a mask. Nevertheless Bug#14509 (and Bug#14350) are still fixed. I successfully tested the test case under "Windows Server 2003" and Wine 1.1.1 with this patch applied. You can find the updates to this patch and test case as attachments. -- Mathias Kosch <info(a)mkosch.de> | http://www.mkosch.de/ RSA Public Key: http://www.mkosch.de/pubkey Mehr Schutz der Privatsphäre: http://www.gnupg.org/

1 0