Patrick's Tips and Tricks

Windows

1. Windows 95 Start button bug
2. Adding shortcuts to the Start menu
3. Windows Explorer shortcuts with starting directory
4. Recovering text files from corrupt floppy disks
5. stdout from Windows programs

Unix

1. Converting ASCII files from DOS to unix format
2. Shell scripts

Java

1. Building Java projects with Microsoft Visual C++
2. Don't forget Mac OS X

Windows

Windows 95 Start button bug

I won a T-shirt for reporting this to a computing magazine in 1996! I suggest you read all of the following before decided to try it out.

Hold down the Ctrl key and press Esc to make the Start button active (Start menu appears - this works on all Windows versions with a Start button and means you don't need to buy a keyboard with a "Start Menu" button). Hold down the Alt key (Start menu disappears but button is still active) and press the space bar. The Start button "system menu" appears. Press the up arrow key to select the Close option and press return - the Start button disppears! (only on Windows 95, on other versions the shutdown menu appears). To log off or shutdown without a Start button, click on the task bar (e.g. where the Start button used to be) and press Alt+F4. You can also move the Start button on Windows 95 by selecting the Move option from the system menu.



Adding shortcuts to the Start menu

Adding items to the main Start menu:

Select the item in Windows Explorer and drag it onto the Start button. See below for how to remove items added in this way.

Adding and removing items from Start menu and submenus.

Right click on the task bar and select "Properties". Select the second tab in the popup dialog box and use the Add and Remove buttons to add and remove items from the Start menu.


Windows Explorer shortcuts with starting directory

It's often useful to create a shortcut to Explorer that starts in a specific directory. You can do this using the /e option of Explorer, e.g. if you're using Windows NT and you want Explorer to start in directory C:\users\jimdandy, change the shortcut command line to:

C:\WINNT\EXPLORER.EXE /e,C:\users\jimdandy

Recovering text files from corrupt floppy disks

If when you try to copy files from a floppy disk you get a read error and it only copies a small portion of the file, find a linux machine and try using the mcopy command, e.g.

mcopy a:/myfile.txt .

This happened to me with the MineSaver source code. When I tried to copy from the floppy in Windows I only got the top few lines of source. When I copied the file using mcopy on Linux, I got the complete file with a line of garbage at the point where the file was truncated on Windows.



stdout from Windows programs

When I started writing Windows programs I didn't know how to use a debugger so all my previous debugging was done using printf statements. I was therefore a bit put out when it seemed that Windows programs didn't have any stdout and printf statements had no effect. My solution was to use the MessageBox function instead, but this gets a bit messy if you need a lot of debugging output to find the bug.

It wasn't until much later (and after I had worked out how to use a debugger!) that I discovered that it is possible to see the output from printf statements in Windows programs. On way of doing this is to run the program from a DOS prompt and send the output to a file, e.g.

myprog > debug.log

Normally you want to see the output from the program as it is produced. You can do this by piping the output to more:

myprog | more

This works, but because stdout is normally buffered you don't see any output until the buffer is full or the program terminates. To get round this problem, you can make stdout unbuffered. This example program demonstrates this technique:



#include <stdio.h>
#include <windows.h>

int WINAPI WinMain(HINSTANCE  hInstance,HINSTANCE  hPrevInstance,LPSTR  lpCmdLine,int  nShowCmd)
{
	int nResponse;
	setbuf(stdout,NULL);  /* Make stdout unbuffered */
	do
	{
		nResponse = MessageBox(NULL,"Press Yes or No for output, Cancel to exit\n",
			"Windows stdout", MB_YESNOCANCEL);
		switch (nResponse)
		{
		case IDYES:
			printf("Yes button pressed\n");
			break;
		case IDNO:
			printf("No button pressed\n");
			break;
		}
	}
	while (nResponse != IDCANCEL);
	return 0;
}


The only remaining problem is that the more command only prints one page of output at a time so you have to press the space bar at the end of each page (although this could actually be an advantage in some situations). One solution is to create a very simple C program that just outputs whatever is piped into it, e.g.


#include <stdio.h>

int main()
{
	while (!feof(stdin)) putchar(getchar());
	return 0;
}

If you called this program winpipe you would then run:

myprog | winpipe

Unix

Converting ASCII files from DOS to unix format

Assuming the file is on a unix machine, there are several ways of doing this. The methods I have used in chonological order are:

1. ftp the file to a DOS machine in binary mode and then ftp it back in ASCII mode. This works but is a bit of a pain if you have to process many files.
2. Write a simple C program to strip out the extra carriage return characters. This is probably the most efficient solution, but I invariably lost the program and its source and had to rewrite it whenever I needed to process a large number of files.
3. Use standard unix tools, e.g. tr:
cat dos.txt | tr -d '\r' > unix.txt



Shell scripts

Shell scripts are just ASCII files that you make executable using the chmod command, e.g.

chmod +x myscript

Most of my scripts for doing batch jobs are based on the following template:



#! /bin/sh

for i in `ls`
do
	echo $i
done

NOTE : The ` character is not a '

This script just prints out the contents of the current directory so is not very useful. A few modifications allows you to convert all the ASCII files called *.txt in the current directory from DOS to unix format:



#! /bin/sh

for i in `ls *.txt`
do
	echo Converting $i
	cat $i | tr -d '\r' > $i.tmp
	mv -f $i.tmp $i
done


"Find in files" script. This searches text files for a given pattern. It searches the given directory and all subdirectories. If no directory is given it uses the current directory.
Example usage:


fif /usr/include/sys/ malloc
/usr/include/sys//malloc.h
56: *   @(#)malloc.h    8.5 (Berkeley) 5/3/95
67: * flags to malloc
289: * The malloc/free primatives used

/usr/include/sys//mbuf.h
230:/* Need to include malloc.h to get right options for malloc  */
231:#include    

/usr/include/sys//namei.h
186: * size a power of two to optimize malloc's. Minimum reasonable

The script:



#! /bin/sh

#Function to explain script usage
usage()
{
        echo "Usage: $0 [directory] pattern"
        echo "  searches for pattern in ascii files"
        exit 1
}

#Check correct number of arguments (1 or 2)
if [ $# -ne 1 -a $# -ne 2 ]
then
        usage
fi

#If only one argument (pattern) use current directory as directory
if [ $# -eq 2 ]
then
        start=$1
        shift
else
        start="."
fi

pattern=$1

#Do the work
for found in `find $start -type f`
do
#       Check if the file is text
        (file $found | grep text) > /dev/null
        if [ $? -eq 0 ]
        then
#               See if the file contains the pattern
                (grep "$pattern" $found) > /dev/null
                if [ $? -eq 0 ]
                then
#                       Found one or more matches print filename and matches
                        echo $found
                        grep -n "$pattern" $found
                        echo
                fi
        fi
done

Finally, the most complex script I have ever written. I used to use this before I used CVS for source control. This script merges two source code trees. You give the source and destination directories as arguments and when it finds source files that differ, it prints out the differences given by diff and then gives you the option of overwritting the destination file, or merging by hand using xdiff. Before using the script you have to configure it for the machine, by adding the name returned from "hostname" to the case statement at the begining of the script.


#! /bin/sh

#
#       script to merge source files
#
#       author patrick@nag-j.co.jp 12.3.99
#
#       Requires GNU diff (for -q) and X11/Motif xdiff
#
#       Set the compare and merge variables in the following case
#       statement to point to these executables on your machine
#
#       If your GNU diff distinguishes between binary and acsii
#       files by saying Binary files x and y differ rather than
#       just Files x and y differ with the -q option, you can set
#       binaryok to 1
#

binaryok=0
diffargs="-wb"
recusivediff="-rq"
mergequiet="-D"
mergeargs="-wb"

#  *** Add your machine to this case statement ***
case `hostname` in
hal )
        compare=/bin/diff
        merge="/usr/sbin/xdiff"
        binaryok=1
                ;;
server )
        compare=/usr/local/bin/diff
        merge="/usr/local/bin/xdiff"
                ;;
extreme )
        compare=/usr/local/bin/diff
        merge="/usr/bin/X11/demos/mgdiff"
        mergequiet="-quit"
        mergeargs="-args -wb"
                ;;
* )
        echo $0: Sorry not configured for this machine
        echo "  If you have GNU diff and X11/Motif xdiff installed add"
        echo "  this machine to the case statement at the begining of"
        echo "  " $0
        exit 1
                ;;
esac

usage()
{
        echo "Usage: $0 [-lnsIm] [-aargs] [-dargs] [-xargs] [-efile] source dest"
        echo "  -I  -- Ignore Imakefiles"
        echo "  -M  -- Ignore Makefiles"
        echo "  -l  -- list mode. Just lists files that differ"
        echo "  -n  -- (nowarnings) Does not warn when file to be overwritten is newer"
        echo "  -s  -- (superuser) Allows source to be overwritten (recover)"
        echo "  -aargs  -- pass args to diff and xdiff"
        echo "  -dargs  -- pass args to diff"
        echo "  -xargs  -- pass args to xdiff"
        echo "  -efile  -- echo everything to file"
        exit 1
}

echotolog ()
{
        if [ $logfile ]
        then
                echo "$*" >> $logfile
        fi
}

printout()
{
        echo "$*"
        echotolog "$*"
}

printnonl()
{
        printf "$*"
        if [ $logfile ]
        then
                printf "$*" >> $logfile
        fi
}

getresponse ()
{
        read ans
        echotolog $ans
}

testcontinue ()
{
        printnonl " Continue anyway (y,[n])?"
        getresponse
        case "$ans" in
                y*)     :
                        ;;
                *)      exit 1
                        ;;
        esac
}

finalcopy()
{
        cp -f $1 $2
        if [ $? -eq 0 ]
        then
                printout $2
                printout Overwritten by
                printout $1
        else
                printnonl Copy failed.
                testcontinue
        fi
}

copyfile()
{
        if [ $nowarning -eq 1 ]
        then
                finalcopy $2 $1
        else
                tail=`basename $1`
                found=`find $1 -name $tail -newer $2`
                if [ $found ]
                then
                        printnonl "***Are you sure?" $1 is newer than $2 "(y,[n])?"
                        getresponse
                        case "$ans" in
                                y*)     finalcopy $2 $1
                                        ;;
                        esac
                else
                        finalcopy $2 $1
                fi
        fi
}



if [ $# -lt 2 ]
then
        usage
        exit 1
fi

nowarning=0
listmode=0
superuser=0
while [ $# -gt 2 ]
do
        case $1 in
        -* )
                options=`expr "$1" : '-\(.*\)'`
                        ;;
        * )
                options=$1
                        ;;
        esac
        case $options in
        e* )
                logfile=`expr "$options" : 'e\(.*\)'`
                        ;;
        a* )
                args=`expr "$options" : 'a\(.*\)'`
                userdargs="$userdargs $args"
                userxargs="$userxargs $args"
                        ;;
        d* )
                args=`expr "$options" : 'd\(.*\)'`
                userdargs="$userdargs $args"
                        ;;
        x* )
                args=`expr "$options" : 'x\(.*\)'`
                userxargs="$userxargs $args"
                        ;;
        * )
        case $options in
        *[!lnsIM]* )
                echo $0: Unknown option $1
                usage
                        ;;
        esac
        case $options in
        *[l]* )
                listmode=1
                        ;;
        esac
        case $options in
        *[n]* )
                nowarning=1
                        ;;
        esac
        case $options in
        *[s]* )
                superuser=1
                        ;;
        esac
        case $options in
        *[I]* )
                skipImake="-xImakefile"
                        ;;
        esac
        case $options in
        *[M]* )
                skipMake="-xMakefile"
                        ;;
        esac
        ;;
        esac
        shift
done

if [ $listmode -ne 1 ]
then
#       check DISPLAY is set and xdiff works
        $merge $mergequiet $0 $0

        if [ $? -ne 0 ]
        then
                printnonl Will not be able to merge files.
                testcontinue
        else
                canmerge=1
        fi
fi

if [ $listmode -eq 1 ]
then
        echo List mode
fi

if [ $nowarning -eq 1 ]
then
        echo Warnings disabled
fi

if [ $superuser -eq 1 ]
then
        echo You are superuser
fi

if [ $skipImake ]
then
        echo Ignoring Imakefiles
fi

if [ $skipMake ]
then
        echo Ignoring Makefiles
fi

if [ $logfile ]
then
        echo Echoing to $logfile
fi

if [ "$userdargs" ]
then
        echo Extra diff args $userdargs
fi

if [ "$userxargs" ]
then
        echo Extra xdiff args $userxargs
fi

echo Comparing files. Please wait...

if [ $listmode -eq 1 ]
then
        if [ binaryok -eq 1 ]
        then
                printout `$compare $recusivediff $diffargs $userdargs $skipImake $skipMake $1 $2 | grep -v "Only in"\
                        | grep -v "Binary files"  | awk '{print $2,$4 }'`
        else
                file1="_"
                for file2 in `$compare $recusivediff $diffargs $userdargs $skipImake $skipMake $1 $2 | grep -v "Only in"\
                        | awk '{print $2,$4 }'`
                do
                        if [ $file1 = "_" ] ; then file1=$file2 ;
                        else
                                $compare $diffargs $userdargs $file1 $file2 | grep "Binary files" > /dev/null
                                if [ $? -ne 0 ]
                                then
                                        printout $file1 $file2
                                fi
                                file1="_"
                        fi
                done
        fi
        exit 0
fi

file1="_"
for file2 in `$compare $recusivediff $diffargs $userdargs $skipImake $skipMake $1 $2 | grep -v "Only in"\
        | grep -v "Binary files"  | awk '{print $2,$4 }'`
do
        if [ $file1 = "_" ] ; then file1=$file2 ;
        else
                continue=1
                if [ binaryok -eq 0 ]
                then
                        $compare $diffargs $userdargs $file1 $file2 | grep "Binary files" > /dev/null
                        if [ $? -eq 0 ]
                        then
                                continue=0
                        fi
                fi
                if [ $continue -eq 1 ]
                then
                        printout \\n\\n\\n
                        $compare $diffargs $userdargs $file1 $file2
                        if [ $logfile ]
                        then
                                $compare $diffargs $userdargs $file1 $file2 >> $logfile
                        fi
                        ($compare $diffargs $userdargs $file1 $file2 | awk '{ print $1 }' | grep ">") > /dev/null
                        noex1=$?
                        ($compare $diffargs $userdargs $file1 $file2 | awk '{ print $1 }' | grep "<") > /dev/null
                        noex2=$?
                        if [ $noex1 -ne 1 -o $noex2 -ne 1 ]
                        then
                                ls -og $file1 | awk '{printf "\n\n< %s\n    Date : ",$NF} \
                                        {for (i=4; i < NF;i++) printf " %s",$i }\
                                        { print "   Size : "$3 }'
                                if [ $noex1 -eq 0 -a $noex2 -eq 0 ]
                                then
                                        printout "          May need merging with"
                                elif [ $noex1 -eq 0 ]
                                then
                                        printout "          Has less code than"
                                else
                                        printout "          Contains extra code not in"
                                fi
                                ls -og $file2 | awk '{printf "> %s\n    Date : ",$NF} \
                                        {for (i=4; i < NF;i++) printf " %s",$i }\
                                        { print "   Size : "$3 }'
                                printout
                                printout Overwrite $file2 "(yes,[no]"\\c
                                if [ $canmerge ]
                                then
                                        printout ,merge\\c
                                fi
                                if [ $superuser -eq 1 ]
                                then
                                        printout ",recover"\\c
                                fi
                                printout ")?"\\c
                                getresponse
                                case "$ans" in
                                        y*)     copyfile $file2 $file1
                                                        ;;
                                        r*)     if [ $superuser -eq 1 ]
                                                then
                                                        printout Overwrite $file1 "(yes,[no])?"\\c
                                                        getresponse
                                                        case "$ans" in
                                                        y*)     copyfile $file1 $file2
                                                                ;;
                                                        esac
                                                fi
                                                        ;;
                                        m*)     if [ $canmerge ]
                                                then
                                                        chmod +w $file2
                                                        if [ $superuser -eq 1 ]
                                                        then
                                                                chmod +w $file1
                                                        fi
                                                        $merge $mergeargs $userxargs $file1 $file2
                                                        $compare $diffargs $userdargs $file1 $file2 > /dev/null
                                                        if [ $? -eq 0 ]
                                                        then
                                                                printout $file1
                                                                printout Merged with
                                                                printout $file2
                                                        else
                                                                printout Not merged
                                                        fi
                                                        chmod -w $file2
                                                        if [ $superuser -eq 1 ]
                                                        then
                                                                chmod -w $file1
                                                        fi
                                                fi
                                                        ;;
                                esac
                        fi
                fi
                file1="_"
        fi
done

Java

Building Java projects with Microsoft Visual C++

These instructions are for Microsoft Visual C++ 6.0 and the Sun Java 2 version 1.3.1 SDK, but they can probably be adapted to other versions of Visual Studio/C++ and other versions of Java or even compilers for other languages.


• A basic project:


1. Create a new empty Win 32 Console project (File | New | Project)
2. Copy the java files to the newly created project directory
3. Add the java files to the project (Project | Add | Files)
4. In FileView, move the files to "Source Files"
5. Specify the custom build for the files (Project | Settings)

In the Project Settings dialog, select all source files and select the "Custom build" tag.
In the command text box (replace C:/Java/Sun/bin/javac with the path to your java compiler):

echo Compiling $(InputName).java
C:/Java/Sun/bin/javac $(InputName).java

In the output box:

$(InputName).class


6. Build the project (Build | Build - F7)
7. Run from the command line


• Additional features:


1. Syntax highlighting

Visual C++ doesn't recognise Java files, but if you right click on a Java file in the text editor and select "Properties" you can change the language to "C/C++" and get comments and most keywords highlighted.
2. Error location

With a C/C++ file if there is an error in the Visual Studio build window you can double click on the line number and the line in the file will be highlighted in the text editor. This doesn't happen with the error output from the Java compiler, because the output is in a different format:

Java Compiler:
test.java:51: undeclared identifier

Microsoft Compiler:
C:\JavaTest\test.cpp(45) : error C2065: undeclared identifier

To make error highlighting work for Java files you can write a program to modify the output of the Java compiler so it has the same format as the Microsoft compiler. If you could use pipes to redirect output in the Custom build command, this program would be very simple. As this doesn't seem to be possible the program is a bit more complex. Here is the code:


#include <windows.h>
#include <stdio.h>

#define PIPE_BUFFER 1024

struct handleInfo
{
	HANDLE process;
	HANDLE pipe;
};

DWORD WINAPI waitForExit(void *ptr)
{
	handleInfo *info = (handleInfo *)ptr;
	WaitForSingleObject(&(info->process),INFINITE);
	CloseHandle(info->pipe);
	return 0;
}


int main(int argc, char *argv[])
{
	PROCESS_INFORMATION  procInfo;
	STARTUPINFO          StartupInfo = { 0 };
	DWORD why,bread,tid,code,i;
	handleInfo hInfo;
	HANDLE writePipe,readPipe,hThread;
	SECURITY_ATTRIBUTES secAttr  = {sizeof(SECURITY_ATTRIBUTES), NULL, TRUE};
	LPSTR lpCmdLine;
	int pos;
	int ac;
	int colon;
	
	StartupInfo.cb          =  sizeof(STARTUPINFO);
	StartupInfo.dwFlags     = STARTF_USESTDHANDLES;	


	if (argc < 2)
	{
		printf("No command to execute!\n");
		return 1;
	}
	if (!CreatePipe(&readPipe, &writePipe, &secAttr, 0))
	{
		why = GetLastError();
		printf("CreatePipe Error : %d\n",why);
		return 1;
	}
	StartupInfo.hStdInput  =  NULL;
	StartupInfo.hStdOutput =  writePipe;
	StartupInfo.hStdError  =  writePipe;
	lpCmdLine = new char[strlen(GetCommandLine())];
	pos = 0;
	for (ac = 1; ac < argc; ac++)
	{
		sprintf(lpCmdLine+pos,"%s ",argv[ac]);
		pos += strlen(argv[ac])+1;
	}
	lpCmdLine[pos-1] = 0;
	if (!CreateProcess(NULL, lpCmdLine, NULL, NULL, TRUE, 
		0, NULL,NULL, &StartupInfo, &procInfo))
	{
		why = GetLastError();
		printf("CreateProcess \"%s\" Error : %d\n",lpCmdLine,why);
		delete [] lpCmdLine;
		return 1;
	}
	delete [] lpCmdLine;
	char buffer[PIPE_BUFFER];
	hInfo.process = procInfo.hProcess;
	hInfo.pipe = writePipe;
	hThread = CreateThread(NULL, 0, waitForExit,
					   &hInfo, 0, &tid);
	if (hThread == NULL)
	{
		why = GetLastError();
		printf("CreateThread Error : %d\n",why);
		return 1;
	}
	colon = 0;
	while (ReadFile(readPipe,buffer,PIPE_BUFFER-2,&bread,NULL))
	{
		if (bread == 0) break;
		for (i = 0; i < bread; i++)
		{
			if (buffer[i] == '\n') colon = 0;
			else if (buffer[i] == ':')
			{
				colon++;
				switch (colon)
				{
				case 1:
					buffer[i] = '(';
					break;
				case 2:
					buffer[i++] = ')';
					buffer[i] = ':';
					if (i == bread) bread++;
					break;
				}
			}
		}
		buffer[bread] = 0;
		printf("%s",buffer);
	}
	if (!GetExitCodeProcess(procInfo.hProcess,&code)) code = 1;
	return code;
}

If you compile this program to create an executable called JavaToVC.exe you then just add this program to the front of the custom build command, e.g:

echo Compiling $(InputName).java
C:/JavaToVC/JavaToVC.exe C:/Java/Sun/bin/javac $(InputName).java



Don't forget Mac OS X

A few tips to make your Java program "Mac friendly"
• Put the menu bar where Mac users expect it to be

Unlike Windows and Unix, all Mac applications use a single menu bar at the top of the screen. To put your Java application at the top of the screen simply add the following line before calling setJMenuBar:

System.setProperty("com.apple.macos.useScreenMenuBar","true");


• Add handlers for the standard Mac help and quit menu items