The "IT DOESN'T WORK!" thread

[mr-mitch]

Thank you. I suppose "" is the 'null string'.

I looked up the PHP manual entry for strings after reading your post and it states this:

NULL is always converted to an empty string.

So that explains why it wasn't behaving as I expected.

I think it's a little misleading that isset() states 'is set and is not NULL' while NULL is "" as a string.

[LDJ]

Hi guys,
For awhile, I've had a Perl xkcd downloader script that I've been using, to save all the comics+ALT text to my computer, and, until recently, it worked. It had been put together from an old script I found on the net+ some help from people, etc.

But, for the life of me, I cannot figure out why it will no longer work.
The problem lies in that, the script SHOULD (and did), find the file on the server for the comic, and save the file, with whatever extension it had. (Most are PNG, but some are JPG)

But now, all files are being saved without an extension, e.g. "MyFile.", instead of "MyFile.png"


I know it's a stretch, but could anyone shed some light on to what the problem might be?

Thanks so much for your help,
Liam.

[phlip]

It would be easier if you posted the relevant bit of your code...

Also, I hope you're using the JSON and not just trying to scrape the HTML...

[LDJ]

Sorry, I wasn't thinking. Also, I've fixed the issue, I had been declaring a variable fairly late in the code, and should have done so at the beginning. After fixing that, it works again.
I've attached the code in case anyone else would like to use this.

Also, I'm sorry, but I'm not using JSON, I have absolutely NO experience or knowledge as to how to use it.

But if anyone wants to expand on this, and incorporate JSON, feel free.

Thanks,
Liam.

xkcd Perl Downloader:

Spoiler:

Code: Select all

#!/usr/bin/perl

use LWP::Simple;
#  use Smart::Comments;

## Objectives ##

#  Download all comics from xkcd.com
#  Ability to download new comics
#  Download ALT text
#  Saved in: ~/Desktop

## Bugs ##

# Files that aren't a png, are being saved as one.
my $ext; #Declare variable to store file extension until file is saved.

# Set Specifics
# etc.
#.
$sitePrefix = "http://xkcd.com/";

## Path to main xkcd directory ##
$path = "$ENV{HOME}/Desktop";


mkdir "$path/xkcd", 0755 or print "xkcd Directory Exists\n";
chomp($path = "$path/xkcd");

$d = get($sitePrefix);
if ($d =~ /http:\/\/xkcd.com\/(\d+)\//) {
    $current = $1;
}

# Obtains all individual comic data
sub getComicData {
    my $siteData = get("$sitePrefix$current/");
    my @data = split /\n/, $siteData;
    foreach (@data) {
        if (/http:\/\/xkcd.com\/(\d+)\//) {
            $current = $1;
        }
        if (/src="(http:\/\/imgs.xkcd.com\/comics\/.+(\.\w{3}))"/) {
            $currentUrl = $1;
            $ext = $2;
            sleep(3);
            if (/alt="(.+?)"/) {
                $title = $1;
                $title =~ s/&#(\d+);/chr($1)/ge;
            $title = "House of Pancakes" if $current == 472;  # Color title on comic 472 with weird syntax
            }
            if (/title="(.+?)"/) {    #title commonly know as 'alt' text
                $alt = $1;
                $alt =~ s/&#(\d+);/chr($1)/ge;
            }
        }
    }
}

chdir "$path" or die "Cannot change directory: $!";
&getComicData();
while ( get("$sitePrefix$current/")){ ### Writing Files $current: $title
    print "Writing Files $current: $title\n";
   # Create directories for individual comics
    mkdir "$current $title", 0755 or die "Previously Downloaded";
    chdir "$path/$current $title" or die "Cannot change directory: $!";

    # Save image file
    $image = get($currentUrl);
    open my $IMAGE, '>>', "$title$ext"
        or die "Cannot create file!";
    print $IMAGE $image;
    close $IMAGE;

    # Save alt text
    open my $TXT, '>>', "$title ALT.txt"
        or die "Cannot create file!";
    print $TXT $alt;
    close $TXT;
    chdir "$path" or die "Cannot change directory: $!";
    $current--;

    # Check for non existent 404 comic
    $current-- if $current == 404;

    &getComicData();
}


# End Gracefully
print "Download Complete\n"

[Yakk]

Google xkcd json gets me this thread:
viewtopic.php?f=11&t=3095&p=1716879
Following the link in that thread:
http://xkcd.com/432/info.0.json
gets me this in a text editor:

Code: Select all

{"img": "http://imgs.xkcd.com/comics/journal_4.png", "title": "Journal 4", "month": "6", "num": 432, "link": "", "year": "2008", "news": "XKCD updates every Monday, Wednesday, and Friday.<br />Previously: <a href=\"http://xkcd.com/374/\">Journal</a>", "safe_title": "Journal 4", "transcript": "[[The man with the hat sits slumped over on a bench, holding his hat]]\nMan with hat: Sigh\n{{alt text: Man, this emo shit was supposed to be for people who didn't have hats.}}", "alt": "Man, this emo shit was supposed to be for people who didn't have hats.", "day": "4"}


which looks utterly trivial to parse, even if one knew nothing about JSON. I might have to spend a short bit learning how JSON escapes things in their quotes (probably backslash based?), and then I could write a five-line perl function to load that into a map (decoded), and the rest of the program becomes trivial.

[headprogrammingczar]

It gets even easier!

[Windowlicker]

Code: Select all

board = new ArrayList<Integer>[31][31];

Error: Cannot create a generic array of ArrayList<Integer>

I wonder why not.. works without the "<Integer>" well enough, and I can sure enough declare "ArrayList<Integer>[][] board;"...
Anyone any ideas why this happens? Like I say the code works fine without it, it's just giving annoying warnings about it.

[headprogrammingczar]

Java can't make arrays of generic types for some bullshit reason. Apparently they expect type safety from generics where there wasn't any for arrays to begin with. Example below:

Spoiler:

Code: Select all

Integer[] test = new Integer[100];
Number[] oops = test;
oops[0] = new Float(3.1415);


Code: Select all

List<Integer> li = new ArrayList<Integer>();
List<Number> ln = li; // illegal
ln.add(new Float(3.1415));


[Rippy]

I'm a bit confused with Haskell type vs value constructors. For example, say I have a type Bar which can be a Foo or an Int:

Code: Select all

data Foo = Foo Int Int
data Bar = Int | Foo

That won't run, citing "multiple declaration of Foo". If you do this, however:

Code: Select all

data Foo = NewFoo Int Int
data Bar = Int | Foo

It runs just fine. It was my understanding, though, that because the situations in which type/value constructors are used, the compiler can always differentiate them, so it's no big deal giving them the same name. It seems awfully arbitrary to have to change the value constructor to something different for just this one type. Surely there's some way of doing this while keeping the same name for the type/value constructors?

[headprogrammingczar]

Right, so what's happening there is a misunderstanding regarding constructors.

When you define a data type in Haskell, the syntax is "data <type> = <constructor> <types of args> | <more constructors...>
What you did in the first is say, "one type Foo has a constructor Foo :: Int -> Int -> Foo. Bar has two constructors; Int :: Bar, and Foo :: Bar". You can get away with Int as a constructor, for the most unintuitive reason imaginable:

Code: Select all

data Int = I# Int#

This is the GHC source for the Int type. There's no Int constructor, so you are free to make Int :: Bar. There is a Foo constructor though, and it is of type Int -> Int -> Foo. When you declare another one, Foo :: Bar, you are contradicting yourself. What you want is not the first or the second, but something like this:

Code: Select all

data Foo = Foo Int Int
data Bar = IVal Int | FooVal Foo

Now you can make a Bar from two constructors: Ival :: Int -> Bar, and FooVal :: Foo -> Bar.

[Rippy]

That makes sense. Has anybody told you lately that you are awesome?

[llamanaru]

I'm working on an SQL project using python. Can anyone see what's wrong with this statement?

Code: Select all

create table PUBACC_CD
(
   Record_Type         char(2)      null,
   Unique_System_Identifier   numeric(9,0)   not null,
   ULS_File_Number         char(14)   null,
   EBF_Number         varchar(30)   null,
   Year_Sequence         smallint   not null,
   Gross_Revenues         money      null,
   Year_End_Date         datetime   null,
   Aggregate_Gros_Rvn_DE      money      null,
   Aggregate_Gros_Rvn_CB      money      null,
   Total_Assets         money      null         
)


Error is specifically

Code: Select all

exceptions.ProgrammingError: (1064, "You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax to use near 'money\t\tnull,\n\tYear_End_Date\t\t\tdatetime\tnull,\n\tAggregate_Gros_Rvn_DE\t\tmoney\t\tnull' at line 8")

Thanks for any help you can give me.

[phlip]

There isn't a datatype called "MONEY", which is where it's falling down. You want one of the numeric types that exist, probably something like DECIMAL(10,2).

[llamanaru]

That would be a problem. It never occurred to me that they'd give me bad SQL definitions... Thank you very much.

[phlip]

Apparently there is a type called "MONEY" in MSSQL... it isn't standard, though. Maybe the notes have been ported from one DBMS to another and they forgot to change it? And didn't test it?

[llamanaru]

That's probably it. The project is to translate some FCC database management stuff originally written in Visual BASIC and MS Access into something more scalable. So it's true, they've been ported and they I forgot to change it and didn't test it. Thanks again.

[Rippy]

I'm afraid I'm having some trouble figuring out State/Reader stuff in Haskell. Wondering if anyone can help me figure out why the following doesn't work:

Spoiler:

Code: Select all

type TileArray = IOArray (Int,Int) Tile

data AppData = AppData {
    fps :: Timer,
    tiles :: TileArray
}

data AppConfig = AppConfig {
    screen :: Surface,
    tileSheet :: Surface
}

getTiles :: MonadState AppData m => m TileArray
getTiles = liftM tiles get

showTile :: (MonadIO m, MonadState AppData m) => Int -> Int -> m ()
showTile x y = do
    tiles <- getTiles
    tile <- readArray tiles (x,y)           --Problem happens here
    let tileIndex = fromEnum $ terrain tile
    screen <- getScreen
    tileSheet <- getTileSheet
    applySurface' x y tileSheet screen $ Just $ clips !! tileIndex
    return ()

I'm trying to pull out the IOArray from the state so that I can read an element from it. However, it gives this error for the line containing readArray:

Spoiler:

Code: Select all

Evsim.hs:97:12:
    Could not deduce (MArray GHC.IOArray.IOArray Tile m)
      from the context (MonadState AppData m)
      arising from a use of `readArray' at Evsim.hs:97:12-32
    Possible fix:
      add (MArray GHC.IOArray.IOArray Tile m) to the context of
        the type signature for `showTile'
      or add an instance declaration for
         (MArray GHC.IOArray.IOArray Tile m)
    In a stmt of a 'do' expression: tile <- readArray tiles (x, y)
    In the expression:
        do { tiles <- getTiles;
             tile <- readArray tiles (x, y);
             let tileIndex = fromEnum $ terrain tile;
             screen <- getScreen;
             .... }
    In the definition of `showTile':
        showTile x y
                   = do { tiles <- getTiles;
                          tile <- readArray tiles (x, y);
                          let tileIndex = ...;
                          .... }

I'm reworking a function obtained from the Lazy Foo SDL tutorials, which originally looked like this:

Spoiler:

Code: Select all

showTile :: (MonadIO m, MonadReader AppConfig m) => Camera -> Tile -> m ()
showTile camera@(Rect cx cy _ _) (Tile tbox@(Rect tx ty _ _) tileType) =
    when (intersects tbox camera) $ do
        screen <- getScreen
        tileSheet <- getTileSheet
        applySurface' (tx - cx) (ty - cy) tileSheet screen $ Just $ clips !! tileIndex
        return ()
 where tileIndex = fromEnum tileType

So originally it only grabbed info from the Reader part (AppConfig); "camera" was part of AppData, but in this case it's passed as an argument. I can't afford to pass the entire array around repeatedly. Now I need information from both AppConfig and AppData, and I can't find any examples to help figure it out. If anyone can see what it is I'm not getting, I'd really appreciate it (I understand I'm being pretty vague). In the meantime I'm reading up more on State and Reader.

[headprogrammingczar]

Code: Select all

    Possible fix:
      add (MArray GHC.IOArray.IOArray Tile m) to the context of
        the type signature for `showTile'
      or add an instance declaration for
         (MArray GHC.IOArray.IOArray Tile m)


This is the fix. Your showTile function doesn't have a class constraint it needs, which it mentions above. Add (MArray IOArray Tile m) to the context of the type signature for showTile.

Code: Select all

showTile :: (MonadIO m, MonadState AppData m) => Int -> Int -> m ()
            |-------------------------------|
                   This is the context

-- what you want
showTile :: (MonadIO m, MonadState AppData m, MArray IOArray Tile m) => Int -> Int -> m ()

Learning the GHC errors takes time, but the errors are always exactly what they say, in my experience.

[Rippy]

yyyyyep that did it (along with a few other slight tweaks).

The problem, honestly, is that I don't quite understand the code. So far I'm finding that it's not so much that Haskell is bad for doing graphics, but that you need a much, much deeper understanding of the language than you do for a comparable graphical app/game in C++ in order to do it.

My attitude right now is to just get the graphics bits to work, even if I don't understand it, so that I can work on the pure code which I do understand for now, and figure out the rest later. There's only so much reading you can do before you have to put something into practice before moving forward.

So, that said, I appreciate the patience!

[headprogrammingczar]

Yeah, all the graphics shit is pretty much its own language anyway, so learning an unfamiliar language in another unfamiliar language isn't going to be very easy. You seem to be having better luck than I did though.

[Rippy]

Eh, I am somewhat stuck again. The main program loop complains when it tries to call my showTile function (the one I added the IOArray class constraint to). loop is of type "ReaderT AppConfig (StateT AppData IO)", and though I forget the exact error, I imagine that's the problem somehow.

I have caved and started reading again though, and I've gotten to monad transformers, which is helping tremendously with understanding what's going on.

[headprogrammingczar]

Transformers are awesome. Not only do they give you the functionality of multiple monads at once, but they also turn into sweet cars.

(More seriously, I suggest looking at the source code for some simple transformers, along with their MonadTrans instance (iirc) to see what lift does)

[tipo test]

MySQL, silly example database:

Code: Select all

create table fruit(id varchar(34));

insert into fruit values (1), (2), (3), (4);

create table person(id varchar(34));

insert into person values (1), (2);

create table eats(fruit varchar(34), person varchar(34));

insert into eats values (1, 1), (2, 1), (3, 1), (4, 1), (1, 2); /* Only person #1 has eaten ALL of the fruits */

I need a query that retrieves all the persons who have eaten all the different kind of fruits . I guess I'd have to use count() and a subquery but It's not working the way I expect...

[phlip]

I can see two options, neither particularly scalable... the direct approach:

Code: Select all

select id from person where not exists(select 1 from fruit where not exists(select 1 from eats where eats.fruit=fruit.id and eats.person = person.id));

Includes unindexed queries on both person and fruit, but should always work.
Alternatively, the counting approach:

Code: Select all

select person from eats group by person having count(distinct fruit) = (select count(*) from fruit);

Contains an undexed query on eats, and also will only work if there every value in eats.fruit is in fruit.id... ie there aren't any dangling references. Which I think is one of the many consistency checks that MyISAM won't enforce. So be careful with this one. This could possibly be solved by adding a where exists() clause, or an inner join to fruit, to filter out any rows with dangling refs.

[tipo test]

Code: Select all

select person from eats group by person having count(distinct fruit) = (select count(id) from fruit);

Awesome, I understand this and it works! Well if you create properly the tables with primary key constraints there shouldn't be the problem you mention no?

About the other way, I don't get it and in any case you're using the value 1 directly aren't you?

Thanks a lot phlip.

[phlip]

The actual select list doesn't matter in an EXISTS() or NOT EXISTS() subquery... only whether or not the query brings out any rows. So it's idiomatic to do something like "SELECT 1 FROM etc" to indicate that the values aren't being used. You could also do "SELECT *" or "SELECT id" or whatever, but it wouldn't make any difference.

"WHERE EXISTS(SELECT anything FROM etc)" is the same as "WHERE (SELECT COUNT(*) FROM etc) > 0", except it knows it can stop running the subquery as soon as it finds a single matching row... it doesn't need to count all of them.

[Kain]

So, I was trying to get a particular class file to run in TeXWorks for a school project that required us to write it in LaTeX, and through a rather convoluted process actually managed to kill one computer and delete the program from the other. In any case, I decided I would replace TeXWorks with TeXlipse, so that I would actually have a working spell checker, etc. The problem is that, while I can get eclipse (helios) to run, and can get eclipse to find the TeXlipse source website, when I go to actually install it I am informed that

Spoiler:

Cannot complete the install because one or more required items could not be found.
Software currently installed: Eclipse SDK 3.6.2.M20110210-1200 (org.eclipse.sdk.ide 3.6.2.M20110210-1200)
Missing requirement for filter properties ~= $0: Eclipse Product Configuration 1.0.0.M20110210-1200 (org.eclipse.rcp.configuration.feature.group 1.0.0.M20110210-1200) requires 'org.eclipse.rcp.configuration_root.win32.win32.x86_64 [1.0.0.M20110210-1200]' but it could not be found
Cannot satisfy dependency:
From: Eclipse SDK 3.6.2.M20110210-1200 (org.eclipse.sdk.ide 3.6.2.M20110210-1200)
To: org.eclipse.rcp.configuration.feature.group [1.0.0.M20110210-1200]

. I am honestly at a loss for what that is trying to tell me...

[TheChewanater]

Code: Select all

template <typename T>
class RotationMatrix
{
    T a, b, c, d, e, f, g, h, i;
    
    public:
        RotationMatrix (const vector3d<T>& eulers)
        {
            T x = eulers.X;
            T y = eulers.Y;
            T z = eulers.Z;
            
            a = cos(x) * cos(z);
            b = -cos(y) * sin(z) + sin(y) * sin(x) * cos(z);
            c = sin(y) * sin(z) + cos(y) * sin(x) * cos(z);
            d = cos(x) * sin(z);
            e = cos(y) * cos(z) + sin(y) * sin(x) * sin(z);
            f = -sin(y) * cos(z) + cos(y) * sin(x) *sin(z);
            g = -sin(x);
            h = sin(y) * cos(x);
            i = cos(y) * cos(x);
        }
        
        vector3d<T> operator* (T xyz[3])
        {
            T x0 = xyz[0];
            T y0 = xyz[1];
            T z0 = xyz[2];
            
            T x = a * x0 + b * y0 + c * z0;
            T y = d * x0 + e * y0 + c * z0;
            T z = g * x0 + e * y0 + c * z0;
            
            return {x, y, z};
        }
}; 

Code: Select all

float xyz[] = {rotation.x * M_PI / 180, 
        rotation.y * M_PI / 180,
        rotation.z * M_PI / 180};
    
    RotationMatrix<float> r (position);
    position += r * xyz; 

I'm trying to make a flight simulator-ish thing. I'd think that this code should make the ship move foward based on its rotation, but it doesn't. Did I make a trivial mistake somewhere, or is there something about transformation matrices that I misunderstand?

[squareroot]

You're taking sin(position.X), Y, and Z? That... doesn't seem right... do you mean to be taking the sin/cos of some angles, like yaw and pitch?

[_Axle_]

@TheChewanater . .

First, Rotation Matrix really doesn't need to be a template. I can't think of any types that would work other than float or doubles. Especially for a game like setting.

Second, You need the original 'facing' of the object, which will be a vector. This vector then needs to be multiplied by the rotation matrix of the correct angles. This will give the new heading direction. Probably want to normalize that new direction and then multiply that by the current speed. This will give how far along that direction, and then you add that to your original position, this will move it ( non-forced based physics movement ).

From your second code sample, you seem to be building a rotation matrix from your current position, shouldn't that be built from xyz or whatever your current angles are? I can't really double check what your rotation matrix is, since usually for games, it is broken into 3 different matrices ( rotation around X, Y and Z ) separately and then just concatenated together before rotating the vector.

[Yakk]

It being a template isn't a horrible idea. You can swap between float and double, or even fixed point.

Building up your rotation matrix from components is a good idea. You only have to compute it once per frame/time slice, which is basically never at modern computer speeds.

Next, note that your code is doing a full matrix multiplication anyhow. So just write a matrix class.

Code: Select all

// does work for us:
template<int n, typename T>
struct Vector_base
{
  T values[n];
  T& operator[](int i){
    Assert(i >= 0);
    Assert(i < n);
    return values[i];
  }
  T const& operator[](int i) const{
    Assert(i >= 0);
    Assert(i < n);
    return values[i];
  }
  Vector_base() {
    for(int i = 0; i < n; ++i) {
      values[i] = T();
    }
  }
};

template<int n=3, typename T=double>
struct Vector: public Vector_base<n, T>
{};

template<typename T>
struct Vector<3, T>: public Vector_base<n, T>
{
  T& x() { return (*this)[0]; }
  T const& x() const { return (*this)[0]; }
  T& y() { return (*this)[1]; }
  T const& y() const { return (*this)[1]; }
  T& z() { return (*this)[2]; }
  T const& z() const { return (*this)[2]; }
};


A really stripped down vector class. I'd also implement +=, and maybe use the CRTP (google it) to have the Vector_base return Vector types from operators like +=.

But I'm a bit of a template geek. In practice, Vector3 is what you usually want.

Then we toss together Matrix:

Code: Select all

template<int n=3, typename T=double>
struct Matrix
{
  struct slice {
    int i;
    Matrix* m;
    slice(int i_, Matrix<n, T>* m_):i(i_), m(m_) {}
    T& operator[]( int j ) const { Assert(j >=0); Assert(j<n); return (*m).value[i][j]; }
  };
  struct const_slice {
    int i;
    Matrix const* m;
    const_slice(int i_, Matrix<n, T> const* m_):i(i_), m(m_) {}
    T const& operator[]( int j ) const { Assert(j >=0); Assert(j<n); return (*m).value[i][j]; }
  };
  T values[n][n];
  static void zero( T& t ) { t = T(); }
  void ForEachElement( std::tr1::function< void(T&) > func ) {
    for(int i = 0; i < n; ++i){
      for(int j=0; j<n;++j){
        func(value[i][j]);
      } 
    }
  }

  Matrix()
  {
    this->ForEachElement(Matrix::zero);
  }
  slice operator[](int i){ Assert(i>=0); Assert(i<n); return slice(i, this); }
  const_slice operator[](int i) const { Assert(i>=0); Assert(i<n); return const_slice(i, this); }
};


which, for the hell of it, has bounds-checking (in debug) operator[][] implemented.

Then we write:

Code: Select all

template<int n, typename T>
Vector<n,T> operator*( Matrix<n, T> const& left, Vector<n, T> const& right )
{
  Vector<n, T> retval;
  for( int i = 0; i < n; ++i )
  {
    for( int j = 0; j < n; ++j )
    {
      retval[i] += left[i][j]*right[j];
    }
  }
  return retval;
}


which gives you Matrix times Vector multiplication, and

Code: Select all

template<int n, typename T>
Matrix<n, T> operator*( Matrix<n, T> const& left, Matrix<n, T> const& right )
{
  Matrix<n, T> retval;
  for( int i = 0; i < n; ++i )
  {
    for( int j = 0; j < n; ++j )
    {
      for( int k = 0; k < n; ++k )
      {
        retval[i][j] += left[i][k]*right[k][j];
      }
    }
  }
  return retval; 
}


Of course, you could have more fun and write Matrixes as Vectors of Vectors, and teach operator* to deal with non-uniform Vector multiplication in such a way that matrix multiplication "falls out". But that is getting silly (and requires a strange default orientation for Matrixes, so probably isn't worth it). (note that any decent compiler will take the above loops and flatten them).

Anyhow, once you have the above, you can just write a few Matrix primitive generators.

Have one "rotate clockwise around z axis". Then a "rotate z axis to y axis" and "rotate z axis to x axis".

From those primitives, you can write "rotate clockwise around x axis" and "rotate clockwise around y axis" using the above matrix operator*.

Now you can generate the arbitrary rotation matrix by simply multiplying the above matrices together.

And why do it this way? Because you can test the pieces without testing the output. You can examine the "rotate around z" matrix, and manually test if it does what you want. The same for the pieces.

Then you can manually test the operator* (which may or may not be right -- the one I wrote above was off the top of my head, it may not even compile).

Once each piece is tested, you can run them all together and have confidence that the result is right.

If you just toss them all together at once, about the only thing you can do is hope that wherever you copied it from was accurate.

[Ptolom]

I'm trying to get ptrace to tell me when a program segfaults but all I can get out of it is the SIGTRAP ptrace sends itself. Does anybody know anything about how signals are supposed to work with ptrace?
I have

Code: Select all

char** subjargv=argv+1;
  pid_t subjpid;
  subjpid= fork();
  struct user_regs_struct regs;
  unsigned long buffstart=0;
  siginfo_t subjsigs;

  if (subjpid==0){
    ptrace(PTRACE_TRACEME, 0,NULL,NULL);
    sleep(1);
    if(execv (subjargv[0] , subjargv)<0){
      fprintf(stderr,"Cannot Launch %s\n",argv[1]);
      return (1);
    }
  }
  else if(subjpid<0){
    fprintf(stderr,"failed to fork\n");
    return (1);
  }

  wait(NULL);
  ptrace(PTRACE_GETSIGINFO,subjpid,NULL,&subjsigs);
  printf("sig=%d\n",subjsigs.si_signo);


But it always gives me signal 5.

[Ptolom]

Ah, I sussed it. You have to tell the program to continue to get to the SIGSEGV. Solutions always come to me shortly after posting here in desperation.

[sourmìlk]

So, I'm reading a bitmap file into an array of bytes using C++, but I keep getting absurdly high amounts for BITMAPINFOHEADER.biWidth and BITMAPINFOHEADER.biHeight.

First, here's how I define BYTE, WORD, and DWORD:

Code: Select all


typedef unsigned char BYTE;
typedef unsigned short WORD;
typedef unsigned int DWORD;
 

and here's how I define the BITMAPINFOHEADER and BITMAPFILEHEADER structs:

Code: Select all


typedef struct
{
    WORD bfType;
    DWORD bfSize;
    WORD bfReserved1;
    WORD bfReserved2;
    DWORD bfOffBits;
} BITMAPFILEHEADER;

typedef struct
{
    DWORD biSize;
    long biWidth;
    long biHeight;
    WORD biPlanes;
    WORD biBitCount;
    DWORD biCompression;
    DWORD biSizeImage;
    long biXPelsPerMeter;
    long biYPelsPerMeter;
    DWORD biClrUsed;
    DWORD biClrImportant;
} BITMAPINFOHEADER;
 

And here's the code I use to read the data into the BITMAPINFOHEADER struct

Code: Select all


FILE *filePtr;
BITMAPFILEHEADERbmfh;
BITMAPINFOHEADER bmih;

filePtr = fopen(filePath, "rb");

fread(&bmfh, sizeof(BITMAPFILEHEADER), 1, filePtr);
fread(&bmih, sizeof(BITMAPINFOHEADER), 1, filePtr);
 

And here's the BMP file.

[phlip]

Have you turned off any structure padding the compiler may be doing? Check the sizeof() values... are they correct? (They should be 14 and 40.)

Many compilers will pad structure fields shorter than 32 bits up to that amount, so that all the values are word-aligned (for speed). Which will make them not match the layout in the file.

What compiler are you using? On what system?

If you're using GCC (or MSVC, but if you were on Windows you'd just use the structures in windows.h) you want this:

Code: Select all

#pragma pack(push,1)
// struct definitions here
#pragma pack(pop)

[_Axle_]

What Phlip said.

I checked your header size in both VS2008 and GCC, both are coming up with 16 byte size for your BITMAPFILEHEADER. The other is correct at 40 bytes.

[sourmìlk]

What Phlip said.

I checked your header size in both VS2008 and GCC, both are coming up with 16 byte size for your BITMAPFILEHEADER. The other is correct at 40 bytes.

This is probably it.

If you're using GCC (or MSVC, but if you were on Windows you'd just use the structures in windows.h) you want this:

I would rather commit seppuku than use MSVC.

Actually, when I don't use the #pragma preprocessor commands, I'm getting values of 16 and 64.. and when I do, I get 14 (which is correct), and 56 (which isn't). When I don't use them, I'm getting 16 and 64.

At least now I know it's the header sizes.

[bittyx]

Are you compiling on a 64-bit Unix system? Because this table shows that a C/C++ long is 8 bytes on that platform, instead of the expected 4 (which exactly explains the discrepancy).

EDIT (ninja'd): To solve this problem, you could use something like uint32_t from <stdint.h> instead of long. The types defined there are exactly the specified number of bits wide, so it's a good way to make sure your program works as expected wherever you compile it.

[phlip]

Ah, that's a good point. You should be using the stdint.h types for something like this anyway.

[sourmìlk]

Are you compiling on a 64-bit Unix system? Because this table shows that a C/C++ long is 8 bytes on that platform, instead of the expected 4 (which exactly explains the discrepancy).

ahah! Yes, that is exactly it. I didn't know about <stdint.h>, so I'll just include that and use those types. Thank you all.