MD5 in C++ is everywhere if you search for it (except for in the standard library or in boost, which isn’t what I would have expected), but I only needed one implementation and I found this one: RSA Data Security, Inc. MD5 Message-Digest Algorithm.
I wanted to figure out how to tell if my code was compiling in Visual Studio or somewhere else (GCC probably), and I found that you can use #ifdef _MSC_VER.
I was having a problem where Visual Studio was saying that “windows.h” could not be found after having installed the Windows SDK v7.1. I found this article which said to update the path in “HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Microsoft SDKs\Windows\CurrentInstallFolder” in the registry. So I did that, and now everything is working.
I did my first operator overloading in C++ today. My code ended up being quite different to the example code that I found, but I think I’m happy with it:
class FileData {
public:
FileData( const string& path );
static bool LessThan( const FileData& a, const FileData& b );
bool operator< ( const FileData& );
unsigned long size() const { return m_size; }
private:
unsigned long m_size;
char m_md5[ 16 ];
};
static bool operator< ( const FileData& a, const FileData& b ) {
return FileData::LessThan( a, b );
}
bool FileData::operator< ( const FileData& b ) {
return FileData::LessThan( *this, b );
}
bool FileData::LessThan ( const FileData& a, const FileData& b ) {
unsigned long a_size = a.size();
unsigned long b_size = b.size();
if ( a_size == b_size ) {
return memcmp( a.m_md5, b.m_md5, 16 ) < 0;
}
return a_size < b_size;
}
I'm not sure when it becomes either useful or necessary to define the operator both as a static and as a member function. I did both because I wasn't sure what was required.
I was curious about the difference between C++ pointers and C++ references, so I searched and found this which says that basically:
I learned how to read binary files in C++ today. My function (which creates an MD5 hash of a file) ended up looking a little different to the example that I learned from:
const int BUFFER_SIZE = 1024;
int length;
char buffer[ BUFFER_SIZE ];
MD5_CTX ctx;
MD5Init( &ctx );
ifstream is( path.c_str(), ios::binary );
while ( is.good() ) {
is.read( (char *)buffer, BUFFER_SIZE );
streamsize count = is.gcount();
MD5Update( &ctx, (unsigned char *)buffer, count );
}
if ( is.eof() ) {
// it's ok, we're at the end of the file
}
else if ( is.bad() ) {
// bad bit is set
cout << "Bad bit is set while reading '" << path << "'." << endl;
cout << strerror( errno ) << endl;
exit ( 1 );
}
else if ( is.fail() ) {
cout << "Fail bit is set while reading '" << path << "'." << endl;
cout << strerror( errno ) << endl;
exit( 1 );
}
is.close();
I was receiving link error LNK1120 in Visual Studio after adding some C code to my C++ project, and the problem was, as I discovered here (and found more info on here), that I hadn’t declared my C code as “extern C”.
So basically I added #ifdef __cplusplus macros to check if it was C or C++ code and if C++ then outputting “extern C {” with a suitable “}” at the end of the file.
Structs and classes can be laid out with padding between the data members for alignment purposes, and this padding doesn’t get initialised unless you specifically zero it out yourself. So if you’re using memcmp and comparing memory at pointers to structs you might have a problem if you’re expecting structs with equal data members to always be equal.
I was forced to lookup Visual Studio’s Compiler Error C2662 because I was getting it in my code. Turns out this happens when your const’s don’t line up. In my case I had const in the client declaration but the member function I was calling didn’t have the const specification.
This little nugget of code from the referenced article shows how to fix up the problem:
// C2662.cpp
class C {
public:
void func1();
void func2() const{}
} const c;
int main() {
c.func1(); // C2662
c.func2(); // OK
}
I’ve been restoring my archives. Basically I have a bit over 1.3TB of data that I’ve tarballed up and stashed on some disconnected SATA disks, and now that I have a computer with the capacity to hold all that data I’m resurrecting the file shares from the archived tarballs. You can see my restore script here:
#!/bin/bash cd "`dirname $0`" data_path=/var/sata2/data tar xf $data_path/1999.tar.gz --hard-dereference > output.txt 2>&1 tar xf $data_path/2001.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2002.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2003.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2004.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2005.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2006.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2007.tar.gz --hard-dereference >> output.txt 2>&1 tar xf $data_path/2008.tar.gz --hard-dereference >> output.txt 2>&1
The restore.sh script creates an output.txt file that lists any errors from tar during the restore process. I then have a set of scripts that process this output.txt file fixing up two types of common errors.
The first error is that the date of the file in the archive isn’t a reasonable value. For example, I had files reporting modification time somewhere back in 1911, before computers. To fix the dates with this problem I run the following scripts:
#!/bin/bash cd "`dirname $0`"; ./bad-date | xargs -0 touch --no-create
#!/bin/bash awk -f bad-date.awk < output.txt | while read line do # note: both -n and \c achieve the same end. echo -n -e "$line\0\c" done
{
if ( /tar: ([^:]*): implausibly old time stamp/ ) {
split( $0, array, ":" )
filepath = array[ 2 ]
sub( / /, "", filepath )
printf( "%s\n", filepath )
}
}
The second class of error that I can receive is that the file that is being extracted from the archive is a hard link to an already existing file, but the hard link cannot be created because the number of links to the target has reached its limit. I think I used ReiserFS as my file system the archives were on originally, and I’m using Ext4 now. Ext4 seems to have limitations that ReiserFS didn’t. Anyway, it’s not big deal, because I can just copy the target to the path that failed to link. This creates a duplicate file, but that’s not a great concern. I’ll try to fix up such duplicates with my pcdedupe project.
#!/bin/bash cd "`dirname $0`"; ./bad-link | xargs -0 ./fix-link
#!/bin/bash awk -f bad-link.awk < output.txt | while read line do # note: both -n and \c achieve the same end. echo -n -e "$line\0\c" done
{
if ( /tar: ([^:]*): Cannot hard link to `([^']*)': Too many links/ ) {
split( $0, array, ":" )
linkpath = array[ 2 ]
sub( / /, "", linkpath )
filepath = array[ 3 ]
sub( / Cannot hard link to `/, "", filepath )
filepath = substr( filepath, 0, length( filepath ) )
printf( "%s:%s\n", filepath, linkpath )
}
}
#!/bin/bash cd "`dirname $0`"; spec="$1" file=`echo "$spec" | sed 's/\([^:]*\):.*/\1/'` link=`echo "$spec" | sed 's/[^:]*:\(.*\)/\1/'` #echo "$spec" #echo Linking "'""$link""'" to "'""$file""'"... #echo "" if [ ! -f "$file" ]; then echo Missing "'""$file""'"... exit 1; fi cp "$file" "$link"
I then checked for anything that I’d missed with my scripts with the following:
#!/bin/bash cd "`dirname $0`"; cat output.txt | grep -v "Cannot hard link" | grep -v "implausibly old time"