True Maybe False

Time off...?

2011-08-24T02:51:00.002+02:00

What is the deal with me and not being able to have a disconnected vacation? If I don't get to check my work email every day I apparently get cranky.

And I've already gotten yelled at twice for not signing off on my time off during these weeks. It's like sugar - there's just to much of that good stuff out there..

"Artificial intelligence? Oh, I saw that movie!"

2011-07-08T08:28:00.006+02:00

I tried explaining the joys of my new found hobby; Robocode, to a couple of my non-hooked friends yesterday. They all just looked at me as if I was from Mars.

"-Program virtual computer robots? Why?" they said

"-Because it's fun of course!" I tried

"-But what do they do?"

"-Well, they compete against each other in terms of targeting, shooting, scanning radars and moving around on a small battlefield." I brilliantly explained. "Kind of last man standing."

"-But why?" they continued

"-Because.. it's fun and quite cool to try and write smart code for gun targeting on moving objects and real-time reactive evasive behaviors.. and.."

"-..."

In the end I just gave up and accepted their nodding and raising of eyebrows as a sign of the fact that I am still a nerd.

Yay! :)

This weeks top observed CPU usage

2011-02-25T10:53:00.003+01:00

It's not seldom that this is what meets your gaze when working with Qlikview. And the best thing of it all is that it's not a bad thing - on the opposite, it just proves the insanely powerful scalability that Qv enables on large data amounts. In this case ~300 million rows of data (even though "rows" doesn't tell you very much; let's just say it's a LOT).

Enjoy!

Gdi+ access violations

2011-02-08T23:00:00.000+01:00

Recently, I had a case where my company's developer application reported intermittent access violations. The repro was done and a crash dump was produced. Let's see what's up.

This dump file has an exception of interest stored in it.

The stored exception information can be accessed via .ecxr.

Thu Aug 19 13:00:39.758 2010 (GMT+2): (1018.e4c): Access violation - code c0000005 (first/second chance not available)

GdiPlus!LZWDecompressor::FHandleNext+0x172:

000007fe`fc12b68a 44885500 mov byte ptr [rbp],r10b ss:00000000`05a0103f=??

Ok, so we are in some Gdi+ library doing some bitmap stuff. Let's see the call stack.

(Truncated output)

0:022> k
Child-SP RetAddr Call Site
00000000`20567450 000007fe`fc128601 GdiPlus!LZWDecompressor::FHandleNext+0x172
00000000`20567490 000007fe`fc126fb2 GdiPlus!GpGifCodec::ProcessImageChunk+0xdb9
00000000`20567c40 000007fe`fc127132 GdiPlus!GpGifCodec::DoDecode+0x2a6
00000000`20567d30 000007fe`fc0c8274 GdiPlus!GpGifCodec::Decode+0xb6
00000000`20567d60 000007fe`fc0c8225 GdiPlus!GpDecodedImage::InternalPushIntoSink+0x38
00000000`20567d90 000007fe`fc0c49da GdiPlus!GpDecodedImage::PushIntoSink+0x41
00000000`20567dd0 000007fe`fc0c4aec GdiPlus!GpMemoryBitmap::InitImageBitmap+0x122
00000000`20567e90 000007fe`fbfe3b04 GdiPlus!GpMemoryBitmap::CreateFromImage+0x90
00000000`20567ee0 000007fe`fbfe535f GdiPlus!CopyOnWriteBitmap::LoadIntoMemory+0xcc
00000000`20567f50 000007fe`fbfe90fc GdiPlus!CopyOnWriteBitmap::LockBits+0x1a3
00000000`20568070 000007fe`fc00c94a GdiPlus!GpBitmap::LockBits+0xc0
00000000`205680c0 000007fe`fc007929 GdiPlus!GpGraphics::DrvDrawImage+0x278e
00000000`20568890 000007fe`fbfd2e83 GdiPlus!GpGraphics::DrawImage+0x369
00000000`205689f0 000007fe`fbfd2f7c GdiPlus!GdipDrawImageRectRect+0x227
00000000`20568ab0 00000001`404ba91d GdiPlus!GdipDrawImageRectRectI+0xac

00000000`20568b50 00000001`404bd863 OurComponent!SomeClass::SomeImageMethod+0x28d [c:\source\somesourcefile.cpp @ 657]

[...]

Ok, so what now? We see our code running fine, but Gdi+ failing. Is it a problem with Gdi+? Well, not really.

Gdi+ has the nasty side-effect of throwing Access Violations as a result of a failure in it's on-demand buffer allocations. To Gdi+, this is probably no biggy, since it will handle the AV, most probably tries a different allocation and goes on as if nothing happened. Now, the developer application in it's own behavior has built-in error handling, but will not capture these access violations (even though Access Violations are never good - this was the exception, harr harr), which means the behavior is that when running a regular release or with no debugger, the exception isn't triggered.

The situation that occurs of course if having a debugger on the application is that it will halt here as part of default behavior. We do know though, that this is not the Access Violation we're after, but for someone only looking for an exception, it's a real dud. The solution for this is to simply disregard first-chance Access Violations in the debugger. There are more neat ways than just drawing them all over the same line, but in this case there's no need.

We had the customer try it out. So at the next debugging session, punching sx into the console gives us all default exception handling by WinDbg.

Truncated output:

0:022> sx
ct - Create thread - ignore
et - Exit thread - ignore
cpr - Create process - ignore
epr - Exit process - ignore
ld - Load module - output
ud - Unload module - ignore
ser - System error - ignore
ibp - Initial breakpoint - ignore
iml - Initial module load - ignore
out - Debuggee output - output

av - Access violation - break - not handled
[...]

Now, we issue a command (sxd) to simply disregard first-chance access violation exceptions, and then show sx again:

(Truncated output)

0:022> sxd av
0:022> sx
ct - Create thread - ignore
et - Exit thread - ignore
cpr - Create process - ignore
epr - Exit process - ignore
ld - Load module - output
ud - Unload module - ignore
ser - System error - ignore
ibp - Initial breakpoint - ignore
iml - Initial module load - ignore
out - Debuggee output - output

av - Access violation - second-chance break - not handled

That cleared the situation pretty good actually, and we could now catch the real exception, which was a "real" second-chance access violation, and fix the bug.

Mission accomplished.

Jump the grenade

2011-02-04T11:28:00.002+01:00

Am I really the only one that gets profoundly happy when friends tell you that their computer are crashing?

I immediately start seeing debugger scenarios where stack reconstruction and other fun debugging activities are in the vicinity - and my gaze stares onto the horizon, picturing yet another success in troubleshooting a malfunctioning device driver or a locked process of some kind. Drewling and gasping, excited and aroused by the mere thought of that golden .dmp file at the end of the rainbow.

Is that really just me? Really?

How large is YOUR hard drive?

2010-10-27T11:52:00.005+02:00

Seeing this when you open Task Manager just makes me feel all warm and fuzzy inside. Just the fact that this insanely powerful server hardware has more RAM than I have hard drive disk space makes the whole scenario almost rediculous. And yes, that is 64 (sixty-four) logical CPU cores and 256 GB RAM. Note that the little raise on the memory counter shows the fairly decent allocated memory amount of.. 11 GB's.

Now, who wants to encode some video?

..and all that stuff

2010-04-15T12:50:00.006+02:00

I just love when humanity shines through in Microsoft Knowledge base articles. Like this one, where someone actually knows that when referring to "all that stuff" - people get it anyways. Especially since the guy writing is a certified Microsoft Most Valuable Professional (MVP).

Read all about "that stuff" here: http://support.microsoft.com/kb/555223

Beast!

2010-04-09T14:03:00.002+02:00

Damn! Belsebub is in my RAM! He didn't even spare the decimals!

Behold - useless hours in the bin! I'll just take those.

2010-01-14T17:01:00.013+01:00

So, one day, I asked to allocate some time to help out doing an evaluation of a virtualization-leveraged platform used to manage software development processes (what a mouthful!).

"-Yes, you might reserve 2 hours per day on this project." said The Boss.
"-Yay!" I said.

One week later I'd managed to get my hands on the bare metal hardware to use as hypervisor platform. Still no luck on the one for the management server, lab software and Postgre database, but that's no worries - I'll just virtualize those on my own hardware, my 8 GB desktop "server" named HAL should manage that just fine. The licenses of the product itself is a later question - we need to get the bottoms running first.

Two busy months later the unused accumulated hours for the project are up to a couple of weeks effective time, and I finally get to action. I download the evaluation license for VMWare ESX 4, skip the MD5 hash verification as usual, and burn the image to a shiny new DVD disc. I can almost hear the 380G5 server buzz with satisfaction as the *nix OS and ESX boots and some ten configuration steps later it rips through those installer packages like there was no tomorrow. Life is good, coffee is still varm, the sound of the server hall is soothing.

Until 26% progress.

And in this case, I tell you; if hell were to be placed on a percentage scale, it would undoubtedly be at 26.

The installer nicely informed me of the following:

"Could not find packageData.pkl file under file:///mnt/source/VMWare/RPMS/"

So, we asked Mr. Google what he had to say about it. A couple of VMWare Community threads about some HWCL issues on Gen 4 of the DL360 servers - this was a G5. Some more about not having Intel VT support at CPU level - that it have. Next was some theories about it was the GUI of the installer that hustled. Well, I did run the GUI, yeah. Let's give it a shot in console mode - but no, same error.

Rebooted, saw some strange message from the HP SCSI controller, and reset the whole RAID array and built a new partition. Reran the installation, prayed to The Almight Coffee Mug and... another message. Still att 26%. What?

"Could not download file "file:///mnt/source/VMWare/RPMS/vmware-esx-cim-4.0.0-1.9.[boring numbers].rpm"

Different errors on the same operation? Not even the new RAID array would seem to cause that. Do the disk have some corruped sectors? And is that mount really on physical disk?

I started to get a little cranky, and got myself some lunch to stop the old brainholder from aching from all the fresh (?) air the server room ventilation throwed on me. An old (?) colleague of mine from IT was at the lunch table and as I spewed my problem on him, he calmly suggested that I would burn a new disc with the ESX image, since the first one might be corrupted, bad quality or just poorly written by the CD writer and/or image writer software. Hence the errors when the installer tried to unpack the binaries.

Hit by his youthful insights (PFY?) I went back to basics and started to verify the image - just in case, and also since I NEVER do that usually. But the MD5 matched. So flip a new disc, press the image (with a new image writer, IMGburn) and returned to the poor, naked hardware once again.

Of course, the installer had to make that dramatic pause at Hell's 26%, which in turn made me freeze like a dick left on Mars. An then *pop*; 27% and onward.

Phu. And that's just the hypervisor. I mean, what are the odds? Anyways, I've got a feeling that this evaluation is gonna be a long one.

The Daily Nerd Quote

2010-01-11T20:22:00.003+01:00

"Once a service contract has been defined, it is implemented in a class, by having the class implement the interface by which the service contract is defined."

I love these mind bending, tounge wrestling and superlative quotes that Microsoft's writers throws around themselves!

And for the record, WCF seems pretty cool.

A lesson in input validation

2009-12-10T16:01:00.009+01:00

Sincerely? Who really encourages their users to use all three special characters of nordic languages in their username? Spotify!

That's like begging for trouble. Why don't you go and make suggestions like

' OR 1=1 --

while you're at it, uh?

That so totally reminds me of this brilliant xkcd strip.

PS. I took the liberty to paste a tasty little image for you non-swedes. DS.

...or else I will be authorized to use deadly force!

2009-12-04T16:56:00.010+01:00

You know, I like Windows Vista. That is of course since I'm one of those people who can actually manage to use a Windows operating system daily without calling Bill Gates a douche and being a total tool of worthless newbies opinions on anti-Windows discussion boards. See, now you got me going again.

However, what I don't like though, is Windows Vista giving me 60 seconds to save my work before my computer restarts.

I mean, if you're gonna off my system, just do it on the spot! Heck, give me a BSOD - just don't leave me hangin'!

I've got 8 GB of RAM and applications worth nearly all that in working set's and you're giving me 60 freaking seconds to save my work? Give me a break. And all because some lsm.exe process in kernel mode screwed up and left a meaningless entry in my Application Event log.

Scheees..

Can Outlook look in your canlendar?

2009-12-04T15:53:00.007+01:00

Having a fetish for useless output and tendencies to gather data like a squirrel does nuts, my instance of Sysinternals DebugView were running in the background of my workstations secondary screen, silently capturing what no-one else is interested in; debug output from running applications.

A normal day, when waiting for data to aggregate and reading some corporate email, I started the usual select-deselect-clicking tics behavior over text in Outlook. Faihtfully, as always, DebugView rattled away some 20 lines of output in the corner of my eye, but this time a interesting thing caught my attention. Something I've not seen before, although I see the same rows hundreds of times every day.

Can someone tell me what a "canlendar" is? And is really the fact that the email item that I click in Outlook is not a "canlendar" so important that it needs to be printed three times each click?

Microsoft! You've got a misspelling and studdering developer lurking around your hoods!