Welcome to The Linux Page

Click to enlarge

This site is a collection of my own work with Linux. Certain things are easy, others take time to decipher and if I ever need to repeat the work (which usually happens!), then I need to remember everything by memory or... have a site with all the steps taken and to take again.

The following are my most recent posts:

C printf(3) for int, long and long long that works every time

Of Note: This only applies to C99, not C++.

I learned something quite interesting today and wanted to keep a note of it.

Whenever I use a printf() with a format such as %ld and try to compile my code on "many" different platforms, I often get warnings on another one.

The fact is that an int, a long, a long long, an int32_t, int64_t all use a different combination depending on your platform and whether you're running in 32 or 64 bit.

So, the answer to this problem are the #define found in the inttypes.h header file:

PRId8
PRId16
PRId32
PRId64

These entries are used after

So many nopw and nopl in amd64 code?!

I have been wondering why gcc adds so many nop instruction in the binary code of my 64 bit programs.

The fact is that code is expected to run faster if properly aligned.

How's that?

A nop instruction does nothing, by definition: No OPeration.

On amd64, the CPU instruction cache (called L1) works by loading 16 bytes at once in the processor decoder. So if you can align your code to a 16 byte boundary, all the better. That way the instructions part of these 16 bytes will be executed at once. When you jump to a non-boundary area, the processor only executes what is left (i.e. say you jump ...

Today I unearthed an old hard drive with Windows XP on it. After a few hours twiddling I finally got the wireless to work on it... although even before that, the svchost application would make use of 99%+ of the processing time.

With just the default System Manager it's hard to find out what really takes time, so I downloaded procexp.exe from the Windows website (DON'T DOWNLOAD A VERSION FROM ANYWHERE ELSE!) and that showed me the tree and thus which tool was using all the processor time.

Surprise! Surprise!

The problem was the automatic windows update. (the

Server Memory Upgrade

I got some new memory for my GA-P55A-UD4P motherboard. It is supposed to support 16Gb of RAM: sweet!!!

So, I got some Kingston which in general the Gigabyte motherboards support. (I had some before.)

I turned off the server and then the powersupply1.

I remove the 2 x 2Gb of RAM and install the new 4 x 4Gb sticks. Make sure it's well sited multiple ...

1. I have one of those power supply you can turn off, there is a switch on it... Without doing so, I can still see some lights on the network connectors meaning that electricity is still running in the machine!

I had a problem for the last 2 days and had a hard time to determine what it was...

Now I know how to resolve such a problem, so I wanted to share my finding! It's very simple, but writing about it here may help me later to save time...

So... I wrote a C++ class for a DLL, this means I want it's public functions to be exported:

class MY_CLASS_API myClass {
public:
    ... // declarations and functions
};

Up to here, nothing strange.

The MY_CLASS_API is a macro defined in some random header file:

#if defined(MSWINDOWS)
#  if MY_CLASS_EXPORTS
#    define MY_CLASS_API

Problem

A few days ago, I made many updates in one of the MS-Access forms I'm working on.

Then, when I tried to open the form I'd get this error...

The Expression On Load you entered as the event property setting produced the following error: User-defined type not defined

This doesn't mean anything to me since I did not define any type... Removing the Form_Load function (commenting it out) made that very error go away, however, since that's the sub-form of another form, it still generates errors when the main form is trying to initialize the sub-form.

Note that it ...

I always wondered why we couldn't just use dpkg to verify that the installed packages have not be tempered with. I know that the feature is hardly safe if the md5's are saved on the same computer, even the same hard drive...

Still, once in a while it's just a manual error that requires me to check that I did do something wrong on my computer.

Under Red Hat the rpm packager offers the -V option: Verify.

Under Debian and thus Ubuntu and other derivatives, the debsums tool has to be installed and used for that purpose.

  sudo apt-get install debsums
  debsums -s <package ...

Got a problem with Qt today... and it took me a little time to figure out again and I'm sure some people have something about it, but I couldn't find the right answer anywhere.

Qt has those objects named QObject. You can derive them that class and to complete the object, you want to add a Q_OBJECT inside the class. Something like this:

  class MyClass : public QObject
  {
    Q_OBJECT
    [...your object definitions...]
  };

That Q_OBJECT is what makes your object visible in the Qt environment (i.e. a Find by name will find your object if you give it a name with setObjectName().

I just upgraded an Ubuntu server from version 9.10 (Karmic Koala) to 10.04 (Lucid Lynx) and started getting a strange error about some package dependencies being unmet. I'm sure I did not intentionally install those packages, so as I was upgrading, I would imagine that they should have been removed automatically if required.

  # apt-get install <some-module>
  Reading package lists... Done
  Building dependency tree      
  Reading state information... Done
  You might want to run `apt-get -f install' to correct these:
  The following packages have unmet dependencies:

Upgrade from Ubuntu 9.10 to 10.04

Today I did an upgrade of a server from 9.10 to 10.04. We were on a server version before upgrading to 9.10 but we could not directly upgrade to 10.04 (working upgrade paths are very specific; see a list here: https://help.ubuntu.com/community/UpgradeNotes )

The most surprising part was the python script at the end.

  /usr/bin/python /tmp/unique-folder/lucid --mode=server --frontend=DistUpgradeViewText

The command line itself is not specifically strange. However, the behavior at the end of the script is a bit strange, mainly because I hadn't see it before. Last time the upgrade was

Unfreeze your Mouse when it Froze

Thank you for visiting my website.

Main Menu

Server Memory Upgrade

Problem