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

AppDB Issues
by Jeremy Newman 08 Feb '08

08 Feb '08

As many of you have noticed we've been having a bit of server trauma this past month. The issue boils down to a bug somewhere in mysqld. We are still trying to find the exact cause of the bug. We do have some leads. I just wanted to do a brain dump on wine-devel here before I go on Vacation for a week. The problem: mysqld will randomly eat up all resources on the system until the server is completely unresponsive to anything but pings. My band-aids thus far: 1. mysql is now running at a nice of 1. This should allow us to at least ssh in and kill it if it gets out of control again. 2. I've removed InnoDB table support to save on memory. Not many tables where using it anyway, and nothing is using transaction support at this time. 3. I've patched the appdb to use memcached for session management. This will reduce some load on the DB as the session table was busiest and slowest table of the appdb. These fixes should give longer uptime, but they are not solving the core issue. While I'm gone, if the server goes offline again, please contact Alexandre or Jeremy White. They can kick proverbial butt of the server to get it going again. Lastly, sorry for the issues. -Newman "We do what we must because we can" - GLaDOS

1 0

Re: More correctly handle the presence of the default ime inputwindow on Ime open and close
by Dmitry Timoshkov 08 Feb '08

08 Feb '08

"Aric Stewart" <aric(a)codeweavers.com> wrote: > diff --git a/dlls/imm32/imm.c b/dlls/imm32/imm.c > index 4acbde3..f3ed6f9 100644 > --- a/dlls/imm32/imm.c > +++ b/dlls/imm32/imm.c > @@ -190,19 +190,17 @@ static LRESULT ImmInternalSendIMENotify(WPARAM notify, LPARAM lParam) > > static void ImmInternalSetOpenStatus(BOOL fOpen) > { > - TRACE("Setting internal state to %s\n",(fOpen)?"OPEN":"CLOSED"); > + TRACE("Setting internal state to %s\n",(fOpen)?"OPEN":"CLOSED"); Wouldn't it be better to change indentation to commonly used 4 spaces instead of a strange mix of 3 and 4 spaces? -- Dmitry.

2 1

Fixing make test - progress update
by Jeremy White 08 Feb '08

08 Feb '08

I believe that we have an easy way to find out the 'state of the art' wrt to make test. That is, my patch + request of a few days ago: http://www.winehq.org/pipermail/wine-devel/2008-February/062303.html resulted in a very nice, easy to digest report http://test.winehq.org/data/20080205/ That data has five test runs and shows clear across the board failures in msi:install, shell32:shellink, user32:listbox, and user32:msg. So, we can quickly and easily collate and visually see the make test failures. And, heck, I'm sure that fixing those failures is now a trivial exercise. Now we just need the will. :-/. Not sure how to arrange that, but I think having make test succeed on all systems should be a mandatory requirement for Wine 1.0, so I think we have to work on solving it sooner rather than later. The todo's as I see them are as follows: 1. Revise winetest to not require the 'make dist' step, so that the function of my patch can be implemented more cleanly. 2. Persuade people to run a script on a regular basis so we always have this data. It's trickier than the average bear because the tests require the 'real' console, so you can't just stick it in a headless cron job. Maybe some xauth magic... :-/ 3. (The hard one) Focus in and fix the issues. I'll work on #1 and #2 (although not for a week or two), if everyone else agrees to do #3 <evil grin>. Cheers, Jeremy

3 3

Re: user32:scroll.c: Compulsory enabling/disabling scrollbars after setting scrollbar`s flags.
by Anatoly Lyutin 08 Feb '08

08 Feb '08

Oh, Sorry! It crashes some another scrollbars. :( -- Best regards Anatoly Lyutin.

1 0

Re: msvcrt: Fix EOF behavior on read from pipe
by Damjan Jovanovic 08 Feb '08

08 Feb '08

On Feb 8, 2008 4:31 PM, Dan Kegel <dank(a)kegel.com> wrote: > Nguyễn Thái Ngọc Duy's patch was almost right, he > forgot a test case, and he neglected to move where > EOF is detected for pipes instead of just removing it. > > Should fix http://bugs.winehq.org/show_bug.cgi?id=11187 > > > > The problem is more general: consoles *also* break on short reads (Java 1.4.1's gdb takes 1 command then quits because it thinks it's the end of file) and I'm sure short read != end of file for ordinary files either, especially in *nix where read() can easily be interrupted by a signal. Damjan

1 0

re: [PATCH] msvcrt: Fix early pipe closing
by Dan Kegel 08 Feb '08

08 Feb '08

On Sat Feb 2, Nguyễn Thái Ngọc Duy wrote: > Pipes are not like other normal files. If we do not read enough > bytes, it does not mean it reaches EOF > ... fixes broken pipe on git-fetch-pack (MinGW port) This is the same problem as http://bugs.winehq.org/show_bug.cgi?id=11187 so it's my fault. Thanks for the information about git-fetch-pack failing, and for the proposed fix. Before we can commit a fix like yours, we need a test case, so I improved msvcrt/tests/file.c to test EOF on pipes more thoroughly. It shows your patch was almost right. I've fixed up your patch and will send it in. - Dan

1 0

kernel32: Implement ReplaceFileW
by luis.busquets＠ilidium.com 08 Feb '08

08 Feb '08

I have read this patch sent in several times and have realised that every time it contains a large number of "goto"'s. I certainly do not like "goto" in structured code. My question to the forum is if it is allowed to use "goto" in wine or if you prefer not to use it. Even when this will make the code larger but it may give a little nore readability,

2 1

Patch tracking system - again...
by Steven Edwards 08 Feb '08

08 Feb '08

Hi, I'm not sure if this would even really be of use with most of the development being done on git anyway, but after re-reading the governance thread when the point of patch tracking was brought up, It got me thinking. last time around it got shelved as normal due to not wanting to disrupt Alexandre's work flow. Thinking about it a bit more I think I've come up with a solution to the problem that should not disrupt Alexandre's flow if I understand git properly. Let me preface this with saying that I don't. At first I thought we might be able to use the sha1 hashes that are the patch names themselves to track if a commit had be accepted or not, but then I guess that would only work if we were pushing and pulling directly in to each others tree's as I see the commit hash differs from the patch hash in the patches I've reviewed. I can only assume this is because of the way git views looks at the entire repo as one chunk of data rather than dealing with single patches. Then as I dug in to it a bit more I noticed the index's did match provided Alexandre did not make any changes before he committed. I was thinking this could be a use for a tracking system. Lets imagine there is this patch daemon that sits out there and reads wine-patches, storing the diff and parsing the index in to a database. This deamon could do a git fetch every 30 mins or so to compare and see if a patch has been merged in based on the index. It seems like we already have most of the tools we need to do it now if someone really wants the tracking system. We know about the patches and we know if its been merged or not, defining rules for accepted or rejected seem pretty trivial to me as we already have the rule, if its not merged in a week then resend. So here is what it would look like as far as work flow goes, someone submits a patch, it goes to this database and up to a website with a listview showing all the patches. Anything not merged yet still in the two week window would show up in the list as white meaning still pending. The further up the list we would have patches that were merged and accepted as is, like the following (showing up in green) http://www.winehq.org/pipermail/wine-patches/2008-February/049719.html http://source.winehq.org/git/wine.git/?a=commit;h=ac61b90f87a2062b99f6f7068… The following patch would show up in yellow as the index differs for the Makefile part of the patch as Julliard made local changes http://www.winehq.org/pipermail/wine-patches/2008-February/049720.html http://source.winehq.org/git/wine.git/?a=commit;h=5bad12c6bc3072a2676f0d9ed… Any other patches that don't get merged would show up in the list as red after the one week window went by. The tracker would assume they were rejected. The cool thing about this is that we could directly link it in to bugzilla so that when a patch is applied the bug could get automatically updated. Now this all hinges on being able to use the diff index to do the tracking. I imagine there might be conflicts if your trying to use that as the only data point as its not as long as a real hash, but assuming your only doing a compare over as period of a few weeks or a month I think it would work. Am I wack? Is there even still the desire to have a graphical patch tracking system? -- Steven Edwards "There is one thing stronger than all the armies in the world, and that is an idea whose time has come." - Victor Hugo

2 1

re: Code quality (was Re: comctl32: Fix invalid syntax)
by Dan Kegel 08 Feb '08

08 Feb '08

Bang Jun-young wrote: >In fact, this is a well known mistake many newbie Win32 developers >make (and fix in minutes). It shouldn't have been in the tree in the >first place if he actually have read the patch. There are a lot of >easily catchable bugs in the tree, for example, potential security >holes like buffer overrun, meaningless comparison of unsigned < 0 (or > > 0), misuse of BOOL vs. HRESULT, misuse of functions such as > strcasecmp(), use of non-portable syntax, etc. Most of them (if not > all) could be filtered out if he have read the patches carefully. >That's the main reason why Wine keeps crashing every time I give it a >try with my Windows apps. I don't think so. In my experience, it's either a missing feature or something more subtle. Outright bugs of the kind you describe are the minority. I think. > Since 1993, Wine has never gotten to the > point where everybody could rely on it for his daily work. It has as > awfully many bugs as Win95. I see something fundamentally > wrong with development process. Nah, Alexandre usually does a good job of screening patches. We're lucky to have such a good maintainer, IMHO. (I say this knowing full well that newbies are put off by how difficult it is to get feedback because I have seen much, much worse maintainers. Imagine if, say, that guy who always yells at the newbies to not paste logs in bug reports was the maintainer. We'd probably lose all our developers in a month.) So you're peeved you don't have commit access. That's understandable, but it's time to get over it. IMHO the single-committer system works pretty well once you get into the swing of it. It actually forces at least one code review on every patch! - Dan

1 0

War on conformance test suite failures!
by Dan Kegel 07 Feb '08

07 Feb '08

Great progress was made at Wineconf last year on getting our test suite closer to always passing on wine, but http://test.winehq.org/data/200801301937/#Wine shows that, according to the results collected from people who run programs/winetest on Wine, twelve DLLs' tests are currently failing. (Let alone the test failures on Windows!) http://bugs.winehq.org/show_bug.cgi?id=9916 also talks about the fact that "make test" shows failures. This situation greatly confuses new developers, who assume that any test failure is somehow their fault. And that's a reasonable assumption, let's make it so! It would make validating a new Wine development environment easy -- it'd be great to be able to say If it doesn't pass "make test" with no errors, something's wrong with the build environment. What's the best way to encourage people to fix these problems? Should we designate some upcoming day, week, or month as a conformance test bug squishing party? Should we offer prizes (or at least notoriety) for the person who does the best job fixing conformance test suite bugs (as judged by Alexandre or his designate)? Or is a simple call to arms enough?

5 8