ReloadSystems

Resetting MySQL password in Linux

michael_cowan — Tue, 26 Oct 2010 15:41:29 +0000

After a clean install of MySQL in Fedora 13 I found myself in the position of being unable to set a root password as one had somehow already been created. The situation of forgetting a password is much more likely, never the less the steps are the same:

Kill the mysql service
[root@host /]# service mysqld stop

Start the service with disabled grant tables
[root@host /]# /usr/bin/mysqld_safe --skip-grant-tables &

Open a SQL command prompt
[root@host /]# mysql

Connect to the DB
mysql> use mysql

View users
mysql> SELECT * FROM `user`;

Reset the user password (’root’ can be any user and ‘newpassword’ any password)
mysql> UPDATE `user` SET `password` = PASSWORD('newpassword') WHERE `User` = 'root';

Quit SQL
mysql> exit

Restart the service
[root@host /]# service mysqld restart

Avoid Dereferencing in Carmacks Implementation of Approximate Roots

michael_cowan — Sun, 08 Nov 2009 11:17:37 +0000

Just about everyone who has even looked into maths optimisation has come across the 'Quake III Magical Inverse Square Root Implementation' as used by John Carmack. For those that don't know it is a 32bit integer magic number implementation of the Newton–Raphson method for finding approximate roots.

The original code from Quake III looks like this:

float Q_rsqrt( float number )
{
long i;
float x2, y;
const float threehalfs = 1.5F;
x2 = number * 0.5F;
y = number;
i = * ( long * ) &y; // evil floating point bit level hacking
i = 0x5f3759df - ( i>> 1 ); // what the fuck?
y = * ( float * ) &i;
y = y * ( threehalfs - ( x2 * y * y ) ); // 1st iteration
// y = y * ( threehalfs - ( x2 * y * y ) ); // 2nd iteration, this can be removed
#ifndef Q3_VM
#ifdef __linux__
assert( !isnan(y) ); // bk010122 - FPE?
#endif
#endif
return y;
}

Okay, so what's the problem? The problem is the line commented as 'evil floating point bit level hacking' - the author didn't like it and I don't like it either. If you compile this code with gcc the following warning will be thrown:

warning: dereferencing type-punned pointer will break strict-aliasing rules

To find out what type-punning is and how to fix it we can turn to the gcc docs:

-fstrict-aliasing
Allows the compiler to assume the strictest aliasing rules applicable to the language being compiled. For C (and C++), this activates optimizations based on the type of expressions. In particular, an object of one type is assumed never to reside at the same address as an object of a different type, unless the types are almost the same. For example, an "unsigned int" can alias an "int", but not a "void*" or a "double". A character type may alias any other type.

Pay special attention to code like this:
            union a_union {
              int i;
              double d;
            };

            int f() {
              a_union t;
              t.d = 3.0;
              return t.i;
            }
The practice of reading from a different union member than the one most recently written to (called ``type-punning'') is common. Even with -fstrict-aliasing, type-punning is allowed, provided the memory is accessed through the union type. So, the code above will work as expected. However, this code might not:
            int f() {
              a_union t;
              int* ip;
              t.d = 3.0;
              ip = &t.i;
              return *ip;
            }
Every language that wishes to perform language-specific alias analysis should define a function that computes, given an "tree" node, an alias set for the node. Nodes in different alias sets are not allowed to alias. For an example, see the C front-end function "c_get_alias_set".

Enabled at levels -O2, -O3, -Os.

It is now clear what the problem is and writing my own implementation becomes trivial:

union _data32 {
long l;
float f;
} data32;
float Math::rsqrt( const float value ) {
// Implementation of the Newton Approximation of roots.
long i;
float x, y;
const float f = 1.5f;
x = value * 0.5f;
y = value;
// i = *(long *) &y;
data32.f = y;
i = data32.l;
i = 0x5f3759df - (i>> 1);
// y = *(float *) &i;
data32.l = i;
y = data32.f;
y = y * (f - (x * y * y));
// Uncomment to calculate sqrt
// y = y * (f - (x * y * y));
return y;
}

- FIN -

496a

Triple Boot Vista, Linux and OSX

michael_cowan — Sun, 18 May 2008 22:41:16 +0000

Recently I bought a new Vaio VGN-SZ and as per usual I immediately zeroed the drive and began creating my patented partition layout. It was here that I paused for a second and thought about my recent need for an Intel Mac. A quick Google later looking at the prices for a Mac Air, and work on triple booting my Vaio was underway. The result of this work is documented below.

NOTE: I started by removing all partitions from the drive, including the VAIO recovery partition. This is not required and if preferred you can skip the first step in part 1.

Software needed (other than Vista and a Linux distribution)

Kalyway Leopard 10.5.1 Installation DVD
SystemRescueCD
A copy of chain0 boot loader

1. Setting up Windows Vista

I always start with a zeroed drive, so the first step is to install Windows Vista. The procedure for installing Vista is very simple and will not be covered here. When installing avoid using the entire drive to leave space for the other partitions required for Linux and OSX.
By default Vista will use the entire drive for itself. To reduce the size of the Windows partition:
- Boot into Sys Rescue, type startx and run gparted.
- Right click the Vista NTFS partition, select Resize/Move and change the size of the partition leaving enough space for Linux and OSX.
- Click Apply to commit the changes made.
While we're modifying the NTFS partition its a good time to create a HFS+ one for OSX:
- Right click the unallocated partition, select New.
- Enter a size for OSX and change the Filesystem to unformatted. Click Apply to make our partition changes. NOTE: We really want this partition to be hfs+ but the version of gparted I was using could not create this, requiring me to use unformatted and perform the next step.
- Quit gparted and enter fdisk /dev/sda in the Terminal. Type p to see a list of partitions and note the number of the unformatted linux one (/dev/sda2 in my case). Enter t to change the partition type, enter the partition number and enter af as the hex code. Enter w to commit the changes and exit fdisk.

2. Setting up OSX Leopard

Restart the computer and insert the Kalyway Leopard 10.5.1 installation DVD. When the Darwin boot loader is displayed press F8 and enter -v. Wait for the installer to load, this can take some time.
Select your language and click past the welcome and license screens.
On the 'Select a Destination' screen click Utilities -> Disk Utility. In Disk Utility select the disk that is greyed out. This is the partition we created earlier for OSX. Click Erase, select 'Mac OS Extended (Journaled)' as the Volume Format and click Erase. Confirm the action and then close the Disk Utility.
Select the new disk added to the 'Select a Destination' screen and click Continue.
Click Customize on the Install Summary. Tick Vanilla_Kernel ACPIPlatform, open Bootloaders EFI, remove BOOT_efi_guid and tick BOOT_efi_mbr. Click Done.
Click Install and wait for the installation to complete.

3. Setting up Linux

Restart the computer and insert the Linux DVD.
Install as normal with one exception; the option to install the boot loader to the Linux partition and not the default MBR must be selected.

4. Preparing the boot options

Boot into Sys Rescue
Activate the Vista partition - Enter fdisk /dev/sda in the bash shell Type p to see a list of partitions and note the number of the one marked as Boot (/dev/sda2 in my case) and the number of the Linux boot partition (/dev/sda4 in my case). Enter a to change the partition boot option then enter the partition number. Repeat the last operation this time toggling partition 1. This activates the Vista partition as boot. Enter w to commit the changes and exit.
Mount the Vista partition - Enter ntfs-3g /dev/sda1 /mnt/windows in the shell.
Copy the Linux boot sector - Enter dd if=/dev/sda4 of=/mnt/windows/linux.boot bs=512 count=1 to copy the Linux boot sector. NOTE: replace the sda number with that noted from fdisk.
Unmount the Vista partition - Enter umount /mnt/windows
Because changes have been made to the NTFS partition Vista will need to recover from the shock. To do this restart the computer, insert the Vista DVD and when asked select Repair and restart. Allow the process to complete and remove the DVD when the computer restarts. Boot into Windows Vista and allow CHKDSK to run if it starts.

5. Putting it all together

Open a Command Prompt in Vista as administrator and enter the following commands to add Linux to the Vista bootloader.
NOTE (replace ID with the number returned by the first command)
- bcdedit /create /d "Linux" /application bootsector
- bcdedit /set {ID} device boot
- bcdedit /set {ID} path \linux.boot
- bcdedit /displayorder {ID} /addlast
Download the chain0 bootloader and copy it to c:\osx.boot
Repeat the bcdedit commands for OSX replacing the description and path to the bootloader

- FIN -

X-Clamp Replacement and Shroud Mod (360 RoL Fix)

michael_cowan — Tue, 14 Aug 2007 13:55:55 +0000

Since my previous attempts to increase stability on my crashing xbox 360 (crashed every 1.5 hours) and subsequent death of my main gaming 360 (RoL - 3 lights of death) I have been reading a lot about the suspected causes of the issues widely seen on the console.

The following is a guide to the work I carried out that has so far proven to be a permanent fix for both of my dead 360's and something I recommend all 360 owners (with no warranty) do, even before the dreaded 'Ring of Death'. Based on the findings of RBJTech, these mods are less evasive, simple and reversible (pretty much).

The following is a list of the parts I used and tools required:

Part Description	Notes
(8) - button head M5 x 20mm screws	Ideally use M5 x 15mm. 4 of the bolts will need to be cut to 15mm
(60) - 5mm washers	These should be approximately 0.9mm thick
Arctic Silver 5 thermal compound	To replace the thermal paste used by the factory
Isopropyl Alcohol solution	Used to remove the old thermal grease. Lighter fluid can be used
Small piece of Aluminum sheet approximately 0.5mm thick	Used to cut a shim for the GPU
Insulation tape and card	Increase and separate air flow in the shroud

XCM unlock kit	Or similar - optionally make your own
T9 Torx screwdriver	Used to remove the heat sink screws
T10 Torx screwdriver	Used to remove the case screws (A T9 will also work)
Long nose pliers	To remove the X clamps
Tin snips	For cutting an Aluminum shim
5mm drill bit and drill	Expanding the heat sink mount holes

1. Strip the Console

The first step is to strip the 360 and remove the heat sinks. Try and work on a large flat surface and I advise the use of an anti static wrist band. XCM have provided a good tutorial for using there unlock kit, so I wont provide too much detail here.

Start by removing the face and top and bottom covers. Split the case and remove the bottom half.
Remove the DVD drive, cables, shroud, fans and RF unit (the front panel with 4 LED's). The plastic cover on the RF unit will need to be removed to reveal the 3rd T9 screw.
Flip the console over. You will be able to see the 8 T9 screws that hold the heat sinks on. Remove these then remove the remaining T10 screws.
Remove the board. This is best done by lifting the front first and letting it slide out.
Turn the board over to reveal the X clamps that need replaced. To remove the clamps I used a pair of long nose pliers and carefully twisted the ends until they came free. This can take some considerable effort but must be done with extreme care. If any of the delicate surface mount components are damaged the xbox will not be easy to repair.
Remove the mount screws from the heat sinks. The clamps, mount screws and plastic spacers will not be needed again.
You can now see the exposed GPU and CPU. To clean the old thermal paste, lighter fluid can be used.

2. X-Clamp Replacement Modification

Drill the 8 heat sink holes to 5mm and remember to clean the case after. To seat the board properly in the case and to avoid bending, washers are used to match the height of the seats pressed into the case. I used 5 washers totaling to 4.5mm but I would say 4.3mm is ideal.
Insert the 8 button screws through the mount holes and hold in place with tape on the back. If you only have 4 15mm screws then use these on the GPU (left) side of the case. Put 5 washers on each of the screws.
Cut a shim measuring 25x15x0.5mm from flat Aluminum. Put fresh thermal paste on the GPU and CPU. Place the shim over the GPU and add more thermal paste.
Replace the board into the case, using the tip of a screwdriver to help align the screws into the holes. Place 2 washers on each of the GPU screws. Place 1 washer on each of the CPU screws. Turn the case onto its side, holding the GPU heat sink in place and carefully turn each screw till it beds. Slowly tighten each screw equally until the heat sink is sitting properly and on tight. Do the same with the CPU heat sink.

3. Shroud Modification

To split the air flow evenly, a piece of card 53x78mm is used. This has a section 12x22mm removed from one corner. To help remove airflow from the DVD drive, a piece of card 90x70mm is used. Hold these pieces in place with tape. A tab will need to be cut into the air flow separator to allow it to fit properly.
Replace the fans and shroud onto the console. Cut a piece of card 25x65mm to increase airflow to the CPU heat sink. As before tape into place but add another piece to hold securely to the heat sink. Replace the cables, DVD drive and RF unit.
Reassemble the case and power on the console. You may need to power cycle a few times before the unit powers as normal. Try running a game and notice that the air flow from the fans is hotter than normal. This is good and shows that the modification is working as more heat is being removed than before. If the issue remains check that the bolts are on tight and making good contact.
X-clamp replacement and shroud modification is clearly better than your mother

- FIN -

Useful Terminology

michael_cowan — Fri, 27 Jul 2007 10:29:23 +0000

This is a list of some of the terminology used in this site that you may not be familiar with. Please feel free to suggest additions for this list...

AUTOMAGICALLY /aw-toh-maj'i-klee/ adv. Automatically, but in a way that, for some reason (typically because it is too complicated, or too ugly, or perhaps even too trivial), the speaker doesn't feel like explaining to you.

TESTICULATING. Waving your arms around and talking Bollocks.

BLAMESTORMING. Sitting around in a group, discussing why a deadline was missed or a project failed, and who was responsible.

SEAGULL MANAGER. A manager, who flies in, makes a lot of noise, craps on everything, and then leaves.

ASSMOSIS. The process by which people seem to absorb success and advancement by sucking up to the boss rather than working hard.

SALMON DAY. The experience of spending an entire day swimming upstream only to get screwed and die.

CUBE FARM. An office filled with cubicles.

PRAIRIE DOGGING. When someone yells or drops something loudly in a cube farm, and people's heads pop up over the walls to see what's going on. (This also applies to applause for a promotion because there may be cake.)

SITCOMs. Single Income, Two Children, Oppressive Mortgage. What yuppies turn into when they have children and one of them stops working to stay home with the kids or start a "home business".

STRESS PUPPY. A person who seems to thrive on being stressed out and whiny.

PERCUSSIVE MAINTENANCE. The fine art of whacking the crap out of an electronic device to get it to work again.

ADMINISPHERE. The rarefied organizational layers beginning just above the rank and file. Decisions that fall from the "adminisphere" are often profoundly inappropriate or irrelevant to the problems they were designed to solve. This is often affiliated with the dreaded "administrivia" -- needless paperwork and processes.

404. Someone who's clueless. From the World Wide Web error message "404 Not Found," meaning that the requested document could not be located.

OHNOSECOND. That minuscule fraction of time in which you realize that you've just made a BIG mistake (e.g. you've hit 'reply all')

McPOO. Entering a fast food restaurant with no intention of buying food, you're just going to the bog. If challenged by a pimply staff member, your declaration to them that you'll buy their food afterwards is known as a McPOO with Lies.

AEROPLANE BLONDE. One who has bleached/dyed her hair but still has a 'black box'.

AUSSIE KISS. Similar to a French Kiss, but given down under.

BEER COMPASS. The invisible device that ensures your safe arrival home after booze cruise, even though you're too drunk to remember where you live, how you got here, and where you've come from.

GREYHOUND. A very short skirt, only an inch from the hare.

MONKEY BATH. A bath so hot, that when lowering yourself in, you go:"Oo!Oo!Oo! Aa!Aa!Aa!".

PICASSO BUM. A woman whose knickers are too small for her, so she looks like she's got four buttocks

Guitar Scale Formula

michael_cowan — Tue, 24 Jul 2007 14:23:41 +0000

Last week I found myself wanting for a simple guitar application containing a number of fairly simple utilities that could replace all the books, charts and little devices each with there own specific task. Being as awkward as I am it wasn't long before I was looking into how to solve these problems myself.

Worse Case
First up I have to find a way for the application to hold guitar scales. At first this seemed like a task of titanic proportions. I figured I would want to store at least 15 scales, in all 12 keys up to at least the 12th fret. The average scale has around 6 notes per key which let me estimate there will be around 35 notes per key in a scale, or a total of 35*15*12 = 6300 notes in the scope I set. Given that each note can be one of twelve (A - Ab), each note can be represented as a byte giving a data storage size of around 6.5k. This doesn't sound too bad, until you consider the scope for error. Storing 6.5k of data is one thing but bug checking it is something quite different.

Typical Design
Luckily this would be a very bad design. An application would not need to store every note in a scale that is visible on the neck, only the notes used in the scale itself which averages out around 6 notes. This cuts the data size to 12*6 = 72 values per scale and 15*72 = 1080 for the scope of the project. A good start but there is a better way.

Scale Formula
This is where the scale formula comes in. The formula works by representing each note in the scale by its relation to the major scale. This would mean that a value of b3 would represent a note a half step lower than the 3rd note of the major scale in the same key. This makes it sound more confusing that it really is and can better be explained with an example.

Example
This example will use the minor pentatonic scale in the key of E. The minor pentatonic scale formula is 1, b3, 4, 5, b7.

First lets look at all the formula and note values, which is based on the major scale (this starting point is the same for any scale or key you want to work out):

formula: 1  b2 2  b3 3  4  b5 5  b6 6  b7 7
notes:   A  A# B  C  Db D  D# E  F  F# G  Ab

Now we need to shift the formula list so it starts on the note of the key we want. In this example the key we want is E. This is shown below:

formula: 1  b2 2  b3 3  4  b5 5  b6 6  b7 7
notes:   E  F  F# G  Ab A  A# B  C  Db D  D#

Now we use the scale formula to get the correct notes for this scale:

formula: 4  5  b7 1  b3
notes:   A  B  D  E  G

The notes left are the E minor pentatonic scale.

Method Comparison

	Typical	Formula
average values per scale	72	6
total values for 15 scales	1080	90

Not only does this mean that a huge number of scales can be stored with very little data but it also means that testing is minimized to checking the formulas only. And because the values are calculated and not stored it means that the application no longer has to limit the number of frets it supports due to storage or testing.

But thats not all; not only does this work for all scales, it can be used in exactly the same way to represent any chord. Very handy when writing a guitar application.

- FIN -

24e3

Properly Centering a DIV

michael_cowan — Mon, 23 Jul 2007 16:36:59 +0000

During the development of a scenic designers web site, I came across something I didn't expect to be an issue. What is required is a div with the following properties:

a set width and height
vertically and horizontally centered in the browser window
correctly displayed on both FireFox and Internet Explorer (these as the browsers most used with the site)
repositions itself as the browser window changes size

The most obvious solution would be to use auto margins. By setting the container of the div using min-width the margins would automagically keep the div centered (well it should but Internet Explorer once again proves that M$ can't read and required a hack using text-align). Unfortunately this only works to align horizontally.

The best solution to this that I could find was to set the absolute position of the div with left and top set to 50%. However this is only half the story because this will set the top left of the div to the center of the screen. To correct for the width and height of the div appropriate margins are set. An example is shown below:

div.container {
position: absolute;
left: 50%;
top: 50%;
width: 600px;
height: 400px;
margin-left: -300px; /* Must be half the width */
margin-top: -200px; /* Must be half the height */
}

- FIN -

Cross Browser Semi-Transparent DIV

michael_cowan — Fri, 20 Jul 2007 14:51:39 +0000

I was recently asked to design a new website for a bikes company. They wanted to use a large background image with a semi-transparent (opaque) div box to contain each pages main content. My first thought was to use filters through CSS. Although filters lie outside the scope of both the CSS1 and CSS2 definitions, the mechanism through which they are called is part of CSS. More recently CSS3 has a new property in an attempt to unite browsers. These properties are shown below:

Internet Explorer              filter: alpha(opacity=50);
Mozilla                        -moz-opacity: 50%;
CSS3                           opacity: 0.5;

For compatibility it is recommended to use all three.

The problem with this is consistency. If a div is created with it's opacity set to 50% using the mentioned methods, different browsers produce different results. Mozilla browsers (rather annoyingly, correctly) apply this opacity to all child elements within the div. So any text, images or even scroll bars are also set to 50% opacity. This is true even if you try to apply an opacity of 100% to these child elements. This is because the 100% you apply to them is 100% of the 50% they currently have. Internet Explorer does not suffer from this and although more desirable, is not consistent with the CSS definition. A new solution needed to be found with the following requirements:

Works on both browsers (IE and Mozilla being the most common browsers visiting the site)
Does not suffer from the 'child inheritance' problem
Has no noticeable negative performance impact

The solution I came up with lay in the PNG image format. I wanted to use a 1x1px image of opacity 30% and tile it across the div as its background. This would give the same result at setting the opacity of the div but without using the CSS properties and having the child inheritance problem. However PNG has it's own issues with support in Internet Explorer which requires a work around supplied by Microsoft. So to set the background image of a div to a PNG file, the following can be used:

Mozilla            background-image: url(box_bg.png);
Internet Explorer  filter:progid:DXImageTransform.Microsoft.AlphaImageLoader(enabled=true, sizingMethod=scale src='box_bg.png');

A final problem with the Microsoft fix in Internet Explorer is that if the standard (Mozilla) method is also included in a div, then the incorrectly rendered PNG image covers the fixed one. To get past this you can exploit Explorers lack of support for attribute selectors. This allows you to place the Explorer fix in the div (which Mozilla will ignore as it is an Explorer filter: attribute) and the Mozilla method in an attribute selector for it (which Explorer doesn't support).
The final div CSS is shown below:

.transbox {
width: 300px;
margin: 10px;
border: 2px solid black;
/* Mozilla will ignore this as is it a M$ filter */
filter:progid:DXImageTransform.Microsoft.AlphaImageLoader(enabled=true,sizingMethod=scale src='semitransparentpixel.png');
}
.transbox[class] {
/* IE will ignore this as it doesnt support styles with [attributes]*/
background-image: url(semitransparentpixel.png);
}

Unfortunately this is still far from a prefect solution. Should Internet Explorer support attribute selectors in the future, then we are back to the background cover up problem. We can only hope that Microsoft sorts out its PNG problems before this ... or, perhaps all browsers will finally unite in their interpretation of opacity in CSS3 and allow the designer to decide if child elements inherit their parents transparency. We can only hope.

- FIN -

2565

Two 360 Cooling Modifications with a Talismoon Whisper Fan, Arctic Silver and a Soldering Iron

michael_cowan — Mon, 16 Jul 2007 14:24:27 +0000

For a short time play.com were selling core xbox 360's for £149.99 delivered - so, sticking to my if it ain't broke, take it apart and fix it philosophy, I took the opportunity to have a second x360 to experiment with more than I normally would with my main gaming machine.

When the console arrived it was immediately identified as having an Hitachi DVD drive with the 78FK bios and I couldn't resist putting to practice the recent work by Maximus. Unfortunately I later found that the 360 suffered from stability issues and tended to crash approximately every half hour (an issue unrelated to the flash). My first suspicion was that heat was the problem (it is also known that the power supply or even live updates have been causing some people issues) and with my warranty still warm and lying in tatters on the floor, I set about looking for third party solutions.

An obvious start is to remove the poor quality thermal compound that Microsoft use and replace it with a high quality counterpart e.g. Arctic Silver. Secondly a simple drop in replacement for the (in my opinion) under powered fans is an easy upgrade to perform. I choose the Talismoon Whisper Fan (in green).

These simple modifications did show a small improvement in both noise and airflow, however provided only a 2-3 degree drop in CPU and GPU temperature when under heavy load.

Next I decided to modify the stock fans to run at their designed 12v instead of the 6v supplied by the 360. This is done by removing the two positive pins in the fan's plastic power block then connecting the wires together and to a 12v supply. There are plenty of good 12v spots on the 360 motherboard to use but be careful when choosing one to make sure the supply is cut when the 360 is turned off - not all of them do.

This modification had two extreme effects. Firstly the noise from the fans is unbelievably loud. And I don`t mean 'the 360 is so loud in comparison to the PS3' loud, I mean Harrier Jump jet loud. And secondly, the temperature drop is equally unbelievable. At idle both CPU and GPU temperatures were way below room temperature and under load there was not much change. more accurate numbers to follow

Unfortunately even this impressive temperature drop in both core and case temperatures had no affect on the stability of my 360. And even more unfortunate for me, it was about this time that my original 360 finally buckled and I got my first taste of the 3 lights of death. I think the modified 360 will be going to GT electronics in Dundee to get fixed, however I`m tempted to try and bring the RoL 360 back using the heat gun technique ... updates to follow.

- FIN -

Left Soft Key Unresponsive After SMS Interrupt on Nokia 5500d

michael_cowan — Mon, 16 Jul 2007 10:14:24 +0000

This is an old issue I hoped I would never have to deal with again as it is caused by a Nokia application distributed with some of their devices. The application in question is the SMS reader which uses a speech synthesizer to read any unread messages when the user holds the left soft key.

During compatibility testing an issue was occasionally reported among a variety of applications regarding the left soft key becoming unresponsive. This is normally a very rare bug to find as all key groups (numerical, directional and softkeys) are treated the same in most application and each group either all work, or all don`t. After some investigation by QA a finger was very quickly pointed in the direction of the SMS reader application.

It did not take long to realize that the issue lay in the calling of keyPressed and keyReleased. A small test application later and it became apparent that the underlying OS was hogging the LSK looking for an extended key press so it could start the SMS reader. The result in the JVM was that keyPressed and keyReleased were being called on the same tick, so in the application nothing happened.

The fix is to detect if keyPressed and keyReleased are called on the same tick when the LSK is pressed. If this condition occurs, keyReleased returns to allow the application to detect and handle the key press. A tick counter is something that many applications have and is handy to fix such bugs. An example is shown:

int lskPressedTick = 0;
protected final void keyPressed(int nCode) {
if( nCode == Device.KEY_LSK ) { // -6 on Nokia devices
lskPressedTick = tickCounter;
}
.
.
.
}
protected final void keyReleased(int nCode) {
if( nCode == Device.KEY_LSK
&& lskPressedTick == tickCounter ) {
return;
}
.
.
.
}

- FIN -