• The Old New Thing

    A quick puzzle about security and synchronization

    • 14 Comments

    This quick puzzle floated past one of our internal discussion groups.

    // in process A
    hEventA = CreateEvent(NULL, FALSE, TRUE, TEXT("MyNamedEvent"));
    
    // in process B
    hEventB = OpenEvent(EVENT_MODIFY_STATE, FALSE, TEXT("MyNamedEvent"));
    WaitForSingleObject(hEventB, INFINITE);
    

    In Process B, the OpenEvent succeeds, but the WaitForSingleObject returns immediately instead of waiting. Explain.

    [Correction: I confused the matter by passing TRUE as the third parameter, thereby creating an event that is initially signalled. Change it to FALSE so that the event is created non-signalled, in which case the WaitForSingleObject would be expected to wait.]

  • The Old New Thing

    Why does Windows XP SP2 sometimes forget my CD autoplay settings?

    • 34 Comments

    It didn't forget them; it's just double-checking with you.

    The developer responsible for CD autoplay in Windows XP SP2 explained it to me. There were two problems with the way Windows XP handled CD autoplay.

    First, when you installed a new program that included CD autoplay capability, many users didn't know where in the UI to go to select that new program as their default CD autoplay program. If they had previously selected a program and ticked "Always perform this action", there was no easily-discoverable way to undo the "always" flag to make the dialog reappeared and allow the user to select the new program instead.

    Second, many programs, upon installation, secretly hacked the undocumented CD autoplay settings in order to set themselves as the default CD autoplay handler, gleefully overriding the user's previously-stated preference. Because these programs egotistically believed themselves to be the coolest most amazing program ever written in the history of mankind.

    In other words, the two problems were, "I just installed this program and I want it to be the CD autoplay program" and its converse "I just installed this program and I don't want it to be the CD autoplay program".

    Windows XP SP2 introduced new behavior related to CD autoplay in an attempt to address these problems: When it sees that a new CD autoplay handler is available, it shows you the CD autoplay dialog one more time. This gives you a chance to (a) pick that new program you just installed, or (b) un-pick that program you just installed (if it was presumptuously rude enough to set itself as your default).

    The first time you insert a CD into your computer after upgrading to Windows XP SP2, you will also get the CD autoplay dialog. This is a "better late than never" dialog to cover for any handlers that were installed before you upgraded to Windows XP SP2.

    What's the moral of the story? Whereas in the old days, you only had to worry about helping other programmers interface with your feature, in the new software landscape, you also have to worry about stopping programmers who are trying to abuse your interface.

  • The Old New Thing

    The 2005 Seattle Chicken Tour

    • 4 Comments

    Mark your calendars for the 2005 Seattle Chicken Tour, scheduled this year for July 16th.

    Seattle Tilths Annual City Chickens and Coop Tour July 16, 2005   10 am — 4 pm
    $10 per family or group of four

    Seattles city chickens owners invite you into their backyards for a first-hand look at raising chickens. Discover the variety of breeds that might be nesting in your neighborhood, learn about raising chickens, see how families integrate chickens into their organic gardening practices, as well as the ingenious coops that shelter these chickens. You will need a car to travel on this self-guided tour of over a dozen backyard chicken broods. Bring the whole family, and your out-of-town visitors, for this unique view of Seattles backyards.

    I am not certain whether advance registration is required, so if you're interested, you should contact Seattle Tilth sooner rather than later.

  • The Old New Thing

    If strncpy is so dangerous, why does Visual Studio 2005 still support it?

    • 45 Comments

    In response to the news that strncpy is so dangerous, at least one person has called for Visual Studio to revoke support for such a dangerous function, considering the continued support for the function grounds for holding the compiler manufacturer liable for any defects in programs compiled with that compiler.

    Well, for one thing, while it's true that strncpy is dangerous if used improperly, it is still a valid function, and my original discussion explained the history behind strncpy and the very specific scenario in which it is still useful. It just so happens that most people don't use the function in the manner it was intended, but instead treat it as a sort of "copy string with a character limit" function, which it isn't really.

    For another thing, just because something is dangerous doesn't mean it shouldn't be supported. Pointers and casts are dangerous, but I don't see them disappearing from C or C++ any time soon.

    Third, support for strncpy is mandated by the C standard. If you removed it, you couldn't call yourself a C compiler any more. (Not to mention breaking compatibility with existing source code that uses the strncpy function. How would you like it if you bought a so-called C compiler and found that it couldn't compile a large class of valid C programs?)

  • The Old New Thing

    Why don't you ever see a rat vomiting?

    • 10 Comments

    Okay, maybe you never wondered why you never saw a vomiting rat, but the intrepid researchers at the Annals of Improbable Research have discovered that there's a good reason, and Anne's rat page will explain in more detail than you probably wanted.

  • The Old New Thing

    Using /LARGEADDRESSAWARE on 64-bit Windows for 32-bit programs

    • 11 Comments

    Probably the biggest advantage of 64-bit Windows is not the larger registers but rather the expansive 64-bit address space. Recall that even when the /3GB switch is set, 32-bit programs receive only 2GB of address space unless they indicate their willingness to cope with addresses above 2GB by passing the /LARGEADDRESSAWARE flag.

    This flag means the same thing on 64-bit Windows. But since 64-bit Windows has a much larger address space available to it, it can afford to give the 32-bit Windows program the entire 4GB of address space to use. This is mentioned almost incidentally in Knowledge Base article Q889654 in the table "Comparison of memory and CPU limits in the 32-bit and 64-bit versions of Windows".

    In other words, certain categories of 32-bit programs (namely, those tight on address space) benefit from running on 64-bit Windows machine, even though they aren't explicitly taking advantage of any 64-bit features.

  • The Old New Thing

    Using modular arithmetic to avoid timing overflow problems

    • 21 Comments

    In an earlier article, I presented a simple way of avoiding timing overflows which seemed to create a bit of confusion.

    The short version: Given a starting time start, an ending time end and an interval interval, the way to check whether the interval has elapsed is to use the expression end - start >= interval. The naive expression end >= start + interval suffers from integer overflow problems.

    To simplify the discussion, let's operate in base-100 instead of base-232. The same logic works, but I think operating in base-100 will be easier to follow.

    Base-100 means that we remember only the last two digits of any number. Consider a starting time of start = 90 and an interval of interval = 10. Using the wrong expression yields end >= start + interval = 90+10 = 100 = 0. In other words, end >= 0 which is always true since end has the range 0...99. As a result, the wrong expression will think that the interval has expired prematurely.

    Using the correct expression, we have end - 90 >= 10. Of the numbers 0..99, the ones that give a difference less than 10 are 90 through 99. Once end = 0, the result is 0 - 90 = 10, which correctly indicates that 10 ticks have elapsed since 90 once the timer reaches 0.

    You can work through a similar mistake using start = 89 instead of start = 90; in this case, the wrong expression becomes end >= start + interval = 89 + 10 = 99, or in other words, end >= 99. This has the opposite problem from the previous case, namely that the expression will fail to detect that the interval has expired once the timer rolls over.

    But why does the end - start expression work? It's very simple: You just have to remember your rules of arithmetic from elementary school.

    (x - c) - (y - c) = x - c - y + c = x - y

    In other words, subtracting the same value from both terms of a difference does not affect the final value. This rule applies even to modular arithmetic (because, as the mathematicians like to say, the set of integers modulo n form an additive group).

    This rule is useful because it lets you delay the overflow as long as possible by subtracting the starting point from all your time markers; it has no effect upon time intervals. Wouldn't it be great if start = 0? Then the overflow won't happen for 100 ticks. Well, you can act "as if" the starting point were start = 0 by simply subtracting start from all your time markers.

    Those who prefer a graphical view can think of time passing as the hands around a clock (which wraps around at 60 minutes, say). When you decide to record your start point, rotate the clock so that the "12" precisely lines up with wherever the hand happens to be. You can now read off the elapsed time directly from your rotated clock. Rotating your clock is the same as subtracting (or adding) a constant to all time markers.

    Of course, this trick falls apart once you have to measure time intervals that come close to the wraparound time of your timer. In our 100-tick timer, for example, trying to measure the passage of 90 ticks is very difficult because there is only a 10-tick window where the inequality is satisfied. If we fail to catch the timer during that window, we miss it and have to wait another 90 ticks.

    So don't do that. In practical terms, this means that you shouldn't use GetTickCount to measure time intervals longer than 15 days.

  • The Old New Thing

    Shocked (shocked!) that patronage exists in Chicago politics

    • 4 Comments

    NPR reported on a startling discovery in Chicago: That government jobs go not to those best qualified to perform them, but rather to those with the best connections. Who'd-a thunk it?

    Shocked by this discovery, the Daley administration vowed to end it.

    The City of Chicago's top lawyer Mara Georges told incredulous City Hall reporters yesterday that the hiring system will now be completely fair, objective, and free of political influence. "City hiring will be on the square, yes."

    The Associated Press notes that "City officials have long denied that political favoritism or patronage had anything to do with who got jobs." The Chicago Sun-Times captures the sentiment quite succinctly: Who knew politics as usual in Chicago is against the law?

  • The Old New Thing

    You can't simulate keyboard input with PostMessage

    • 15 Comments

    Some people attempt to simulate keyboard input to an application by posting keyboard input messages, but this is not reliable for many reasons.

    First of all, keyboard input is a more complicated matter than those who imprinted on the English keyboard realize. Languages with accent marks have dead keys, Far East languages have a variety of Input Method Editors, and I have no idea how complex script languages handle input. There's more to typing a character than just pressing a key.

    Second, even if you manage to post the input messages into the target window's queue, that doesn't update the keyboard shift states. When the code behind the window calls the GetKeyState function or the GetAsyncKeyState function, it's going to see the "real" shift state and not the fake state that your posted messages have generated.

    The SendInput function was designed for injecting input into Windows. If you use that function, then at least the shift states will be reported correctly. (I can't help you with the complex input problem, though.)

  • The Old New Thing

    When is x/2 different from x>>1?

    • 25 Comments

    Everyone "knows" that the following pairs of expressions are equivalent:

    x*2 ≡ x<<1
    x/2 ≡ x>>1

    Too bad they aren't.

    In the C language standard, there is no requirement that the internal representation of signed integers be two's complement. All the permissible representations agree for positive numbers, but negative numbers can have different representations. If x is negative, then x*2 and x<<1 are quite different on a sign/magnitude system.

    However, Win32 requires a two's complement machine, in which case the first equivalence x*2 ≡ x<<1 is indeed always true.

    Of course, the compiler is free to recognize this and rewrite your multiplication or shift operation. In fact, it is very likely to do this, because x+x is more easily pairable than a multiplication or shift. Your shift or multiply-by-two is probably going to be rewritten as something closer to an add eax, eax instruction.

    As for the second so-called equivalence, the C language specification originally did not specify whether division of a negative number by a positive number rounds towards or away from zero, but in 1999, the specification was revised to require rounding towards zero. Furthermore, the result of a right-shift of a negative value is unspecified, so the expression x>>1 has an unspecified result if x is negative.

    Even if you assume that the shift fills with the sign bit, The result of the shift and the divide are different if x is negative.

    (-1) / 2 ≡ 0
    (-1) >> 1 ≡ -1

    The moral of the story is to write what you mean. If you want to divide by two, then write "/2", not ">>1".

Page 368 of 444 (4,431 items) «366367368369370»