This was a rough morning, all right! Especially since it was so difficult to insert an ampersand (&) in that freakin’ label!

So the problem is that & is used for the mnemonic function (you know… in case you want the File menu to pop up when you press the ‘F’ key on your keyboard, you would name the File menu as ‘&File’).


Therefore, having a label text called ‘a&b’ would require:

1. disable the menmonic function: UseMnemonic = False

2. change the label’s text to a&&b


I admit it’s kind of lame to have a post about this, but you just might wonder one day why your label isn’t listening 🙂

Sometimes, just sometimes, it’s very frustrating to search for something that isn’t there, isn’t it? 🙂

I was trying to make a tutorial on how to use the built-in setup and deployment system that VS05 offers in the form of Setup and Deployment projects, and the result was that I was ‘hanging’ at the customization part, specifically at changing things like the ‘default company name’ tag. Why is was so difficult? Well, basically, because it just wasn’t there.

After googling around, I even found out that the tags can be changed using some key inside the HKLMSOFTWAREMicrosoftWindows NTCurrentVersion, but this is totally unacceptable, of course (to be honest, I didn’t even succeed in that method either).

So I’ve been looking and looking for about 1 hour for those impossible-to-find tags, when I finally decide to dock the Properties windows and surprise: there they are.

Take it like this: right-click-ing the Setup project created inside the solution brings you to the Setup Property Pages windows, which contains a set of completely different properties than the ones listed in the Properties toolbox. This is frustrating and confusing, since almost everything in VS works different: right click-ing a component in Solution Explorer toolbox (or even Catalog Item toolbox) will either pop the Properties toolbox or pop a windows with the mostly the same properties listed in the Properties toolbox…

So, whenever you are looking for changing the default tags from a setup project, remember to check the one and only: PROPERTIES TOOLBOX. 🙂

Alex 🙂

coming up next: how to create an installer with a professional feeling, sparing the $1000+ required for InstallShield or InstallWise.

This is one of those days, when you do one thing, you compile, everything works perfectly, but for some reason, the second time you compile, nothing works any more :).

Platform Builder 6 has a very complex sequential build process. In theory, is has four main steps:

  1. the compilation phase
  2. the sysgen phase
  3. the release copy phase
  4. and the make run-time image phase

Each of these steps uses a very large number of small files containing module and file definitions that will be included in the final image, registry keys and values for the run-time image created during a cold boot, RAM file system directories, files and links, database to be included in the object store of the run-time image, created also during a cold boot and locale-specific definitions (strings), that will replace visible texts to a user.

In conclusion, lots of files, which lots of values. Therefore, lots of possible vulnerabilities.

Now, as I mentioned in the beginning of the post, this is one of those days, when a WinCE6.0 process works perfectly, until boom! No modification done to the image, and while building a second time, I get lots of strange errors (ok, I admit, I did add some insignificant registry keys, but there’s no way a registry key could affect missing .tmp files used during a build process).

After a lot of digging up and re-re-re-re-re-re-re-build, I’ve realized it’s all got to do with my antivirus. Even though I pray for health and wealth for its developers when I go to sleep, I do admit it has a drawback: scanning the opened files used by an application will delay application responsiveness and therefore, the build process.

Apparently Kaspersky isn’t related in any way to Platform Builder (unfortunately), so I had to manually detect which are the executables that run most often in the build process of a WinCE 6 O/S image and create an exclusion for them. Since most of the anti-viruses has such a drawback, I’ll list them so you can add them yourself to your anti-virus’ exclusion list as well:

  • Platform Builder addlib (<WinCE installation folder>publiccommonoakbini386addlib.exe)
  • Platform Builder Res2Res Resource Copy Tool (<WinCE installation folder>publiccommonoakbini386res2res.exe)
  • Platform Builder Build File Filter Tool (<WinCE installation folder>publiccommonoakbini386cefilter.exe)
  • Microsoft Resource File to COFF Object Conversion Utility (<WinCE installation folder>sdkbini386cvtres.exe)
  • Microsoft Linker Stub (<WinCE installation folder>sdkbini386editbin.exe)
  • Microsoft Program Database (<WinCE installation folder>sdkbini386mspdbsrv.exe)
  • Platform Builder Program Maintenance Utility (<WinCE installation folder>sdkbini386nmake.exe)
  • Microsoft Incremental Linker (<WinCE installation folder>sdkbini386x86link.exe)
  • Platform Builder PBSpawn Utility (%Program Files%Microsoft Platform Builder6.00cepbIdeVSPBSpawn.exe)
  • Microsoft Visual Studio 2005 (%Program Files%Microsoft Visual Studio 8Common7IDEdevenv.exe)

Addid this files should help you in case you keep getting random BLDDEMO errors, which don’t really say much of an error in the error or log file…

So, in conclusion, there are some executables that would like to open a file, but the anti-virus kicks in, scans the file and executable will stop with an error for not being able to open (read ‘find’, instead :-)) the file. So what you should do is to add a collection of executables as exclusions to the anti-virus, thus not scanning the files that those executables open.

That’s all for now.


P.S.: You might find that xcopy or copy will also be used (doh!). Adding them as an exclusion is however not such a bright idea, since most viruses will use either of them to populate themselves through the drives…

Guess what! The goodies have arrived 🙂

The GPS project I will soon start working on is about to… be started :). Today the courier has finally arrived with the GPS Evaluation board and the GPS receiver module, which rocks :). Surely, there wasn’t much time to test everything, since it have only passed like 20mins since I’ve received the first pieces of the puzzle. So far, I have the GPS-08334 Evaluation board (this is actually optional, but is great – in my opinion – for starters) and the EM-406A SiRF III Receiver. I’ve ordered them from (I can totally recommend it, even if it’s 10% more expensive than other online electronic shops). Pictures were taken with my HTC HD2 (aren’t they awesome?).

IMAG0031 IMAG0033

Pieces (of the puzzle) yet to receive: the Vortex embedded board.

Just for fun, I’ve tested the module using the SiRFDemo application, and everything worked like a charm! I’ve connected the eval board to the development station (which is currently an IBM T60 laptop) using the USB connection available due to the evaluation board. For the final project, I intend to use RS232 only.

So what’s next? I’ll write my own NMEA class for the communication with the GPS receiver module (again, over RS232).

To be continued…


Here’s to my BFF, Vlad!

Words are not adequate enough in this difficult times he’s facing! Vlad, lean on your friends for strength, love and happines! And most important, always remember that we’ll be there for you!

With lots and lots of sadness in my heart,


I’ve just realized there is no working solution in C# on how to programatically select cells across a dataGridView.

What I needed was a quick, simple and easy solution on the next behaviour: a user clicks a cell (any kind of cell: may it be a textBox cell, a checkBox cell, a comboBox cell etc.), the cell gets selected, the user clicks another cell, both the new cell and the previous cell are selected, the user click the first cell again, only the second cell is selected.

How I’ve done this? Well, first, it’s important to keep in mind that the events occur like this: CellMouseDown, CellClick, CellContentChanged, CellMouseUp, SelectionChanged. Sure, there are so many other events as well, but it’s important that you use only the CellContentChanged and the SelectionChanged events. Moreover, be aware, that if you try using the CellClick event, you won’t get this desired behaviour :).

What’s more to add, please keep in mind that you will have the MultiSelect property set to True. It would be also nice to have the SelectionMode set to CellSelect, so you don’t need to take care of several exceptions that could occur if you select the entire row/column by clicking on the column header or row header.


– first, declare a System.Collections.ArrayList arrayList = new System.Collections.ArrayList(); This is where you will save a list of all of your selected cells. Sure, it would be nice to save the SelectedCells collection of the DataGridView, but that collection is obviously read-only. Otherwise, I probably wouldn’t even have this post :P.

– second, in the SelectionChanged event, do this:

foreach (DataGridViewCell arrayElement in arrayList)
arrayElement.Selected = true;

And in the CurrentCellChanged, do the following:

if (dataGridView1.CurrentCell != null)
label1.Text = (dataGridView1.CurrentCell.RowIndex + 1).ToString() + ” ” + (dataGridView1.CurrentCell.ColumnIndex + 1).ToString();
if (arrayList.Contains(dataGridView1.CurrentCell))
dataGridView1.CurrentCell.Selected = false;
dataGridView1.CurrentCell.Selected = true;

This is basically it. Now, there is a thing about this code. The thing is that is you click on a column header, in order to re-sort the column, you’ll have a rather strange “error”: your last selected cell will get de-selected. If you re-sort the column again, the cell will be selected again. The fault of this occurrence is due to the way the events are called. Basically, when you click the header, the CurrentCellChanged is called for a first time, since you clicked the header (so, dataGridView.CurrentCell will be null). Afterwards, the CurrentCellChanged is automatically called again, because current cell’s position is moved (so the event is called for some reason…).

There are several ways to get along with this thing; I’ll let you find your own solution, since:

  1. I for one, needed NonSortable columns
  2. I needed the column headers hidden

Alex M.

It’s not that I didn’t know this for such a long time, but today I’ve just realized that there are so many questions regarding automatic application launching.

So, leaving the startup process aside, once WinCE has loaded, the system loads all the applications specified in the HKLMInit registry key, with the values of LaunchXX. The XX stand for a integer value between 00 and 99 (this means that 1 should be coded as 01). The XX values actually represent the sequence that applications are launched (so 00 will be launched first, 01 second etc.). It is good to keep in mind that the system will not wait for an application to ‘fully’ start. This means that it will generate an interruption for each and every application in the Init key, in a specified sequence, without creating a delay or a timeout between any two consecutive applications in the sequence. This is good since it means that application launching cannot create any kind of system hangs (at least in theory).

However, it is sometimes necessary to create a dependency between applications (application MyShell.exe cannot start – in the developer’s point of view – without running Shell.exe, since, for example, it is using some variables generated by Shell.exe). Therefore, a DependXX value can be used, where XX should match the value in the LaunchXX key where the dependecy is specified.

LaunchXX contains a value of the REG_SZ type which must be the name of the program that needs to be launched, such as MyShell.exe, without the parameters. Again, the XX value determines the load order.

DependXX contains a value of the REG_BINARY type, which enables you to determine the dependency of applications on other applications during the load by specifying which applications must be loaded prior to the application specified in the corresponding LaunchXX key. The indexes of XX applications that a given application is dependent on are specified as a lists of words with a reverse byte order. It is important to keep in mind that a word is (as usual) 2 bytes long.

The application specified in the Init registry key must inform the system that it has loaded successfully and the dependent applications can be loaded by calling the SignalStarted() function with a parameter that it is passed to by the system as a command line parameter
during the load.

An example of the Init registry key is shown below:

“Depend50”=“hex:14,00, 1e,00”

In this case, the shell.exe application launches first, then it loads the Device Manager (device.dll), which depends (in the example) on shell.exe, then it loads gwes.dll, which is dependent on the Device Manager, and finally, MyShell.exe, which depends on both the Device Manager (device.dll) and gwes.dll.

Normally, setting up the startup options is something you would do on OS image building. However, in case you can’t do this before building the image, and considering you are using a Hive-based registry, it is ok if you use a Registry Editor or a Remote Regustry Editor (like the one over KITL) and update the registry ‘manually’.

Word of advice: always be sure to double-check the launch sequence! It is possible that certain BSP features or drivers are already in the sequence and required by the your OEM. Using the same XX value twice in a sequence may cause system crash or system ‘hang’. In case you want to auto-launch only one application (your main application for the system), it is probably best to assign it the 99 value (in case it’s free :-)).

That should be it for now! Comments are very welcome!


I have had some trouble changing the ScrollBar’s width in an application I’ve been asked to modify. Obviously, the first thing to do was to search for a Size property for the panel’s scroll bar. Just as expected, there wasn’t any. What could I do more? Surely Google (or Bing…) would have the answer!

But surprise! No good answer on the first 6 pages that Google showed up. So I started checking the Components toolbox again, and there it was: VScrollBar.

It is probably the best way to have a larger scroll bar in your application, since this component has a Size property. Linking the component to the form or panel you want to scroll is as easy as enabling the component’s Scroll event and writing down

panelName.VerticalScroll.Value = vScrollBar.Value;

Now, if you want to be 100% sure about how the scroll bar looks like, simply set the component’s Maximum property to the difference of the panel’s length (should I call it height), and the portion of the panel that is not showed (the one that should be scrolled for…).

In the end, I suggest you use a panel, for easier and safer usage of the vScrollBar.

Hope the post is clear enough. If not, don’t hesitate to ask your question in a comment.