Fun with IT

Force.com IDE v23 updated again

2012-01-20T07:25:00.001-08:00

Eclipse shows 23.0.2.201201091635 is available now. I have 23.0.1.201111041306. What I'm looking for the new update to fix include:

App freeze when parsing testing results
Unable to expand custom fields under Standard Objects in the metadata browser (Add/Remove)

Stay tuned...

Crashing Move Player

2012-01-19T20:15:00.000-08:00

This is a follow-up post in a sense, to the one about getting Lucid to work on an old Dell X300. With the tip provided in the Wiki about re-enabling KMS, I was able to get it running. However I quickly ran into another problem: the Movie Player (Totem 2.30.2) bombed out when attempting any AVI file. It's a complete crash, killing X as well as the whole display - I couldn't even get to any console. Naturally it felt like still the Intel 8xx issue. So, maybe it's time to come clean.

I read the Wiki again, and it looks like the more complete fix would be either upgrading the Intel driver, or just upgrading the back-ported kernel 3.0. So I tried the less dramatic one - upgrading the Intel driver. It's a simple straightforward install (you gotta love Apt!). Once that's in and rebooted, I saw some improvement, but not enough. Totem still crashed, but hurting no one else but itself.

Only one thing left to try: the 3.0 kernel. Another painless apt-get - how could replacing the heart of the OS this easy? Anyway, once rebooted, all looked well! Three cheers to the Wiki people.

Ubuntu 10.04 upgrade caused blank screen upon rebooting

2012-01-09T12:32:00.001-08:00

I'd have to say the upgrade process 10.04 (from either 8.04 or 9.10) has been excellent for me. It had worked flawlessly for me from a powerful quad core 12 GB desktop to a meager netbook powered by Cyrix C7-M before I encountered this problem. This happened on a Dell Latitude X300, a reliable machine served as light-duty server for me for years. The upgrade went fine until after restarting, where I was greeted by a blank screen shortly after seeing the splash screen. It turned out to be an issue with the graphics controller inside the old 855PM chipset (https://wiki.ubuntu.com/X/Bugs/Lucidi8xxFreezes). The Workaround A suggested in the wiki (re-enabling KMS, Kernel Mode Setting) worked for me. I just dropped into root shell and ran

echo options i915 modeset=1 | tee /etc/modprobe.d/i915-kms.conf update-initramfs -u

The next rebooting went through without a hitch.

Conga Composer endpoint URL

2012-01-05T07:51:00.001-08:00

Usually it should be https://www.appextremes.com/apps/Conga/PointMerge.aspx, which I think is what's listed in the documentation. You could also use https://www.appextremes.com/apps/Conga/PM.aspx, which seems to work the same except that it doesn't do the multi-select (SelectTemplates=1). I noticed that when trying to modify a single template selection screen to add the multi-select... caused some head scratching before I saw the PM.aspx at the end.

Note that PM.aspx is the page redirected to even when you start with PointMerge.aspx. Not sure if there's an advantage of using PM.aspx directly.

MS Word: "... should be opened as read-only unless changes to it need to be saved. Open as read-only?"

2011-11-18T13:20:00.001-08:00

I got this message when opening a Word document, which I just cleansed with the Hidden Data Removal tool from Microsoft. Not sure why it'd flag the file so, but it apparently did. Here's how to correct it (sourced from this thread):

Open the file - make sure not in Read-only mode
On Tools menu, click Options.
Click Security.
De-select the "Read-only recommended" check box, and then click OK.
Click Save.

Confused Conga Composer

2011-11-17T07:33:00.001-08:00

First of all, Conga Composer is a great document generation tool on the Salesforce.com platform. Probably the best, in my opinion. But it does have some odd issues, like everything in the world. The most odd one is this issue of not able to accommodate two entry points in the same browser session. So if you try to run Conga off both your production org and a sandbox org, which is quite common for a developer/admin, it totally gets confused and becomes unable to retrieve templates. You might see "No Templates available" on the template selection screen, and an odd error like "can't open E:\templates\v0000010345" when actually trying to merge.

Fortunately I usually have different browsers open for different orgs, so that's not an issue for me. I only noticed it once when I was trying to test IE compatibility for multiple orgs. It looks odd because you'd think the transactions should always be identified by SFDC session IDs and instance endpoints, which has nothing to do with the client launch point. Besides, exposing the server info in error messages is never a good idea. :)

Force.com: Testing Apex scheduleable class

2011-09-08T07:48:00.000-07:00

If the execute() method just inserts a batchable run, should I expect after Test.stopTest() the batch is completed? From my last try, it doesn't seem to be the case. I might be wrong, but couldn't find any documentation. The example given in the Apex Developer's Guide was about action done directly in the Execute() call, not the type of "second-degree asynchronous" run I was trying. I have to say my need isn't that unique though - I could imagine a lot of batchable scenarios could become scheduled this way.

Salesforce.com: Change Set unable to deploy Master-detail objects

2011-08-26T13:03:00.001-07:00

The detail object just keeps failing with

"Cannot set ControlledByParent on a CustomObject without a MasterDetail relationship field"

However, the same object deploys just fine with Eclipse (presumably Ant directly too). So, I guess this has to be chalked up to Change Set?

Saw similar report here (see comment by griffin).

SFDC dashboard redirect: to dash or not

2011-08-15T11:51:00.001-07:00

If you want your SFDC dashboard to click through to something else other than the underlying report, an alternative URL can be specified. If the URL is another report in SFDC, the preferred way is to have /reportId as the URL. Using a server name free URL makes refresh to sandbox work without additional effort.

Sometimes you may simply user reportId as the URL. This will work when viewing the dashboard, because that's basically a relative path. However you need to remember that this WILL BREAK DOWN if the dashboard gets snapshot on the Home page. That's because the URL for Home is servername/home/home.jsp, which renders the relative path to an incorrect location. So, I'd almost use the dash version (absolute path) at all times to avoid this subtle issue.

Understanding the OpportunityHistory object

2011-08-11T13:20:00.001-07:00

From Force.com documentation, you'll learn that every record there is a change to one of the following fields on Opportunity: Amount, StageName, CloseDate, Probability, or ForecastCategory. However it's not obvious to see how a stage change (or any change) is recorded (expressed in records). I eventually figured out that two time fields on the OpportunityHistory records are the key:

CreatedDate	The time when this change happened
SystemModStamp	The time this record becomes history (the time when a new OpportunityHistory record following this one is generated)

So for instance, if you need to figure out what the previous stage was for a particular stage, find the OpportunityHistory record with the stage name first. From that record use the CreatedDate and locate a record with the exact value in its SystemModStamp. The stage name in that record would be the previous stage.

What's on the built-in Opportunity History report looks like values derived from the OpportunityHistory object instead of values coming directly from it.

Bluetooth blues

2011-08-09T20:51:00.000-07:00

Bluetooth is a technology meant for short distance (except for hacking of course). I learned that fact when trying to retrofit my desktop with BT based wireless K/M. Since the desktop isn't BT-enabled, I picked up a USB dongle out of the thousands floating out there. With a Dell PU705 Travel BT Mouse and a Dell GM952 BT keyboard, I was in business in short order. However I quickly discovered that if I plugged in the dongle in one of the back USB ports, the mouse became pretty jerky in motion. If it's in a front USB port, it worked OK. Not wanting to occupy any of the two precious USB ports on the front long term, I tried remedying it by swapping out the old battery in the mouse, hoping it's the weak battery causing reception issues. No improvement.

Because of the size of my desk and the arrangement of the stuff on it, back ports are about 1 foot farther away to the mouse than the front ones. And one foot might as well be another continent in the world of Bluetooth! I guess the dongle's power isn't that strong, and the mouse is really intended to pair with a laptop that's usually an elbow away. Combined they're powerless (literally!) facing the 5 feet of chasm between them.

Frustrated but refused to extend my poor wrist just to make it work, I decided to bring the port closer. After locating a powered USB hub (Belkin) in the basement, I connected and put it next to the monitor, then plugged in the dongle. That's about 1.5 feet away from the mouse. Man what a difference that made - the mouse movement is now smooth like butter! Does the power of the strength decrease proportional to the square of distance? I can't remember my college physics anymore, but apparently it's just REALLY significant.

For anyone out there struggling with a shaky Bluetooth connection: getting closer can be the easiest solution!

Good check list for Sites page getting "Authentication Required"

2011-06-06T11:42:00.001-07:00

I was researching the issue and found this good check list on this thread (actually mentioned by the asker), which helped me to pinpoint the cause. The discussion thread was about, and eventually solved by something else... but still want to thank for the valuable background information provided. So, in short, if you encounter "Authentication Required" when trying to access a Force.com Sites page, you should check:

The object in question (used by the Standard Controller, for instance) has read, create, and edit checked for the guest user
The field level security is set to Read Only at least (or better, Edit if allowed) for everything accessed
The guest user has access to VF Pages used
The guest user has access to Apex classes used (direct or indirectly called)
There are no relationship fields on the custom object
The Controller is not using "with sharing" modifier
Any error that may be caused by the page, since if there's an error, you get hit with the same "Authentication Required" page.

"This report can no longer be edited or run..." Not so, pal!

2011-04-08T06:19:00.001-07:00

"This report can no longer be edited or run. Your administrator has disabled all reports for the custom object, or its relationships have changed."

This is really a dreaded error message, if you happened to do something like changing a lookup field to master-detail on an object, then try to use reports associated with it. A tough spot to be in. I was quite confused though, when I saw it yesterday when trying to construct a dashboard. After thinking long and hard and concluding that I'd not committed no such offenses, I set out to find the cause.

The dashboard component was based on a new custom object, which will be the target for snapshots. Nothing special there. No lookup or master-detail fields on it. The object had been deployed, and I can view the underlying report just fine. Well, there's one thing I'd forgotten, it turned out: the dashboard is a dynamic one. And the new object has been opened up to any profile. That means no one can really see that dashboard component other than folks who can "view all". That's not good in production of course... but I'm not quite there yet. The thing has to be reviewed by admins first before opened up to anyone.

Looks like I need to at least allow access to the object by at least one profile explicitly before I can do anything about it in dashboards.

Next time I'd like a more accurate error message please. Well, I guess that's everybody's wish for all kinds of situations...

Number field is always created as Decimal (internal type)

2011-02-10T13:44:00.000-08:00

For all custom sObject in Salesforce.com. Just learned this today:
http://boards.developerforce.com/t5/Apex-Code-Development/Integer-Double-or-Decimal/m-p/202075/highlight/false#M34823

So if you want to use the value in those fields in code for functions as non-decimal argument, a casting member function from Decimal (intvalue, longValue, doubleValue) is always in order, since Decimal sits right at the highest level of the implicit conversion hierarchy.

Useless Isnull() for formula

2011-02-10T11:59:00.001-08:00

That's right, or to be gentler, next to useless in Salesforce.com formula. Let's recount:

Can't be used for TEXT, Rich Text: use Len()=0 instead
Can't be used for Address: use AddrFieldName=''
Can't be used for lookup (ID) fields: use isblank() instead - I just realized it today
Can't be used for pick list: use ='' as well

That's a large chunk of things you may want to check for null condition... Why couldn't they have a more flexible, "overloaded" isnull()?

Client side vs. server side on Visualforce

2011-02-02T06:14:00.000-08:00

What I needed to do was to disable a drop down box based on the selection in another drop down. Fairly common requirement. The controller drop down starts as disabled, so I set disabled property in , and used JavaScript from onchange to toggle disabled in HTML DOM from the controlling drop down. All sounded fine and dandy and the UI worked as it should, until I realized that the form never returned anything from the controlled drop down no matter what its status was.

It looks like the client side JavaScript did nothing to enable the controlled element. As far as the server was concerned, it's a field it can discard. I didn't do any HTML/JS debugging to confirm that, but it certainly felt that way. Once I switched the controlling mechanism to server-side processing/AJAX, it started to work as intended.

The only drawback with the AJAX way was that it's a bit slower. Action from the client-side Javascript was instant. With AJAX and its server round trip, it took about one second for the controlled drop down to change status.

How to override "Log a Call" - the easy way

2011-01-24T07:57:00.001-08:00

"Log a Call" is a standard button you'd see on the "Activity History" related list on any enabled Salesforce.com object page layout. Since it's "Log a Call", the subject is naturally set as "Call". We often get asked if we can put in things like "Phone - Outbound", because when it comes to wording certain things, some business people tend to be quite specific. Well, it's a simple request without a straightforward answer. I've seen people overriding the standard button with a VF page that redirects. I'd like to not code if I can, even though coding very much pays my bills. :) My solution is to use a custom button on Task, List Button type, labeled "Log a Call", shown in the same window without bars, and just a plain adapted URL from the standard button with merge fields, such as:
/00T/e?title=Phone%20-%20Outbound&who_id={!Contact.Id}&followup=1&tsk5=Phone%20-%20Outbound&retURL=%2F{!Contact.Id}
Depending on where it's invoked and what the Subject it needs to have, you can just tweak the parameters and the merge fields. Nice and simple, even an admin can do it.

Apex Class Security: Subtlety

2011-01-04T13:26:00.001-08:00

The "Enabled Apex Class Access" section on a Salesforce.com Profile seems to be straightforward. You can limit whether the profile has access to the methods on Apex classes. Recently I discovered it's not that simple.

I had a profile that had nothing enabled under Apex Class. So you'd think those users don't get to invoke things on those classes. Not so. They would be able to start the VF pages, which displayed just fine. So at least many getters on the controller classes ran just fine. However action methods would fail silently. You can invoke, but nothing comes back (in my case it's outbound calls). Even more intriguing thing is, if you enable debugging on the user, the whole thing would start working magically. And it would stay that way even after the debugging is turned off.

Once I had all classes enabled for the profile, this behavior disappeared. I'd love to do some controlled experiment on that but didn't have time. Just wondering if anyone noted similar issue...

The Internet is full!

2010-12-11T11:51:00.000-08:00

Relax. It's not that bad yet, but may well be if you believed what you saw on this screen:This rather frightening (or silly, depending on how you look at it) message came up more than once when I tried to use the Go-go in-flight Internet service on some of the Delta Airlines planes. This season (Nov. 20th, 2010 to Jan. 2nd, 2011) Delta and Google had a good idea together on how to promote their wares - letting people use the in-flight Internet service for free, "Happy Holidays from Google Chrome!" What a nice thing to do! No one really wanted to think about the pesky issues like bandwidth and throughput until people started to take on their offer. I guess Delta only had two choices: let users languish with the connection slower than a crawl, or ration it so for those who're lucky enough to be on, the connection is still useful. No good option here, and they chose the second one for whatever reason they had.

I encountered the screen about half of the time I tried. With that kind of rejection rate, it becomes harder to tell if this is a promotion or an effort to turn people off. Call me naive, but someone needs to take some heat for this negligence, or there is no justice in business. :)

Getting Input Method to work on Lucid

2010-12-05T19:49:00.000-08:00

I installed the Chinese language when installing the OS, but it's not available when I tried to use it. Here are the steps I went through to get it:

Checked "Language Support" under System > Administration. It said "Some languages are not installed completely". So, let it finish as prompted, and verified the languages I wanted are indeed there.
Switch the keyboard input method system to "ibus" in "Language Support". The default was "lo-gtk" for some reason.
Still, no joy when trying "Ctrl-Space". So, checking System > Preferences > IBus Preferences. It said "ibus daemon not started". Let it start as prompted.
There seemed to be no additional input available in IBus Prefs other than English. Added what's needed.
Now Ctrl-Space brought up the Panel icon, but still no input UI.
Just out of curiosity, checked the ibus daemon process and noticed there's a defunct ibus-daemon process. Considering the battle of killing a defunct process, I simply restarted - gasp, bad PC habit! :)
After logging in, rejoice!

Case matters with SFMT

2010-11-18T07:07:00.001-08:00

While a big portion of SFDC is not case sensitive, its standalone Ant-based tool IS, every bit of it. Whenever I stumpled on that, the failure messages weren't that obvious on the reason. For instance, if I got the object name's case wrong, it would complain about field names on the object not in package.xml. Yeah, big help here! I know, I know, XML is case sensitive... but it's still inconvenient.

Two SFDC tips helpful for me today

2010-11-09T14:33:00.001-08:00

Show/Hide stuff on VF page with AJAX
http://forums.sforce.com/t5/Visualforce-Development/Basic-Show-Hide-pageBlockSection-not-working/m-p/168722
Basically the rule is to always have the rerendered region bigger than the stuff that has conditional visibility. If they are the same size, you may have the problem of things unable to become visible, because the rerendering never works due to the id not being there the first place.
Create checkbox list on the fly
http://blog.jeffdouglas.com/2009/01/13/apex-search-with-checkbox-results/
This is common for searching-then-processing. However the approach isn't necessary for simple situations where you just need simple labels. You can just use the built-in SelectOption class and <apex:selectCheckboxes>.

The perils of ImplicitParent

2010-11-02T08:35:00.001-07:00

Luckily we're not talking about lost biological father or shared custody. Those could be worse battles, but this one isn't easy to deal with either. If you have an elaborate sharing scheme set up in Salesforce.com with roles and sharing rules in Salesforce.com, and you feel very safe about it, you may be in a surprise when accidentally exposed - by ImplicitParent. That's a rowCause in tables like AccountShare. The Reason in the documentation (APEX Dev Guide) for the rowCause is "Associated record owner or sharing", which isn't quite clear to understand what it means. This link may explain it better. Basically that indicates the Account record is being shared out to users who have access to Opportunities or Cases on the account by ownership or sharing rules. Now you see the danger?

I had just such a case happened to me. A business unit is isolated by sharing model and rules so their records can't be seen by other users. Then an admin imported some opportunities under that unit's master Account record with himself as owner. Shortly after his role was changed for testing things - that change added 4 ImplicitParent rows to the AccountShare for the master Account record, because other than himself, there're sharing rules defined for the new role to give more access to others. All of a sudden the master Account unit and all its Contacts children with sensitive data were exposed to a whole lot new audience (Contacts are "Controlled by Parent"). Luckily all of this were caught in a dry run. Scary!

Setting up SVN service on Windows XP

2010-11-01T07:06:00.001-07:00

Had to do it to tie me over for some proejcts and here are what I found out:

This is a good guide to follow on setting up a simple Windows SVN server.
One thing to keep in mind is that using SC command on Windows, you need to be careful about escaping the double quote if you have a path that requires it, or better yet, just choose a path that doesn't need quoting. If the quotation marks aren't done appropriately, you end up with incorrect path in the registry, and the service can't be started with the error 1053 ("The service did not respond...")
svnserve.exe can't use mapped drive for repository. Use local directory. Otherwise 1053 again.
Remember there should be no leading space in any conf file lines, and there needs to be space on both sides of any "=". Otherwise you get the cryptic "option expected" message when issuing commands.

"Will work for Droid X"

2010-08-27T11:28:00.000-07:00

Damn, I was almost going to pimp myself out on the street with a sign like that. That's how hard to get one of those these days. Droid X went on sale since July 15th and I wonder how many have actually got it? I figured it's not a "new" smartphone anymore so it shouldn't be that hard to pick up one. Well, not so fast, Bob. I first put in an order on Amazon Wireless on Aug. 25th, since they priced it $20 below Verizon. Boy was I surprised to see the estimated delivery date: Oct. 8th.

There's no way I can wait that long. So I went to talk to a Verizon store, hoping to get one quickly from the source. The friendly guy there said their shipping date is much closer, but... still at Sep. 10th. Droid 2 and HTC Incredible are definitely in better shape, available the same day.

I guess depending on who you're asking, Verizon/Motorola has either never learned from Apple's lesson, or just taken a page right out of Apple's playbook. In any case, I'm definitely not the only drooling customer out there.

With this piece of empirical evidence, now it's pretty easy to see the strength of the Android platform.

Force.com Migration Tool on Ubuntu (Lucid)

2010-08-24T13:38:00.000-07:00

Since the tool is based on Ant, I apt-get the ant package. The Guide on the tool talked about putting the jar file into the lib directory of ANT_HOME. Well, I don't know where the ANT_HOME (not found in env), let alone locating its lib. Eventually I found the lib dir with some Googling. It's /usr/share/ant/lib - well, should be obvious enough to me... With the jar file in place and config files done (build.xml, build.properties, package.xml), the tool did run successfully to do the things I needed.

Excel WebApp: 0, Google Docs: 1+

2010-08-15T19:27:00.000-07:00

Tried to use a .xls file in Excel WebApp on Sky Drive, and I got "Unable to load the requested workbook". It's a very simple file, a few columns of numbers and texts, nothing big.

Google Docs, with its minimalist interface, worked this one fine - just like so many others I've done.

Cloud is a killer, isn't it?

Apex Explorer for Force.com: using newer API

2010-08-04T11:20:00.000-07:00

It's a handy tool but the newest version available uses API 8.0. Can't remember when that one came out. :) So I tried to change the Options to use newer APIs. The newest version I can get to is 16.0. Beyond that it pops a "null reference" error and is unable to retrieve object list. So there's no change for anything new in API 17, 18, and 19, and I suspect a lot of things in versions after 8 don't get used either.

ERROR - VERR_SUPDRV_COMPONENT_NOT_FOUND

2010-07-29T07:34:00.000-07:00

This probably started happening to VBox after I let through a kernel update – usually I did “keep current menu.lst” when asked by Update Manager. The driver it complained about is the network, which caused it to be unable to attach to eth3 in bridged mode. The good thing is, after following the very helpful suggestion in the error message – run /etc/init.d/vboxdrv setup as root – it resolved itself in about a minute. That’s much easier to do than VMWare Server, which requires a complete reconfigure.

I don’t remember this happening to my VBox before though, when it’s still Sun Virtual Box and it was running on Jaunty. I always let kernel update to be applied then.

SCIM Chinese input method not showing

2010-07-28T07:06:00.000-07:00

It only brought up US Keyboard, English/European, and Raw Code. Before my laptop was upgraded to Lucid, I think it was working fine. I checked Language settings and all necessary languages are installed. So I sought suggestion from Canonical support, and they asked me to try installing scim-chinese package. That did it. As usual, a spot-on solution. I have to say Canonical's support, along with Fonality's, are the best teams I've ever encountered (Dell's a close third) among the dozens I had experience with. Both are open-source vendors. What did that tell you? ;)

PC5750 on Lucid

2010-06-16T19:40:00.000-07:00

I couldn't get the EVDO card to work under Jaunty on my Latitude E5400. Now I've upgraded to Lucid on E6510, and PC5750 is a piece of cake... well, almost.

After I inserted the card, I got a bunch of evbug.c messages in dmesg, and no ttyACM0 to use. A little search got me to blacklist evbug. I plugged in the card again, and noticed there's an option "Enable Mobile Broadband" in Network Manager. So I did it and chose Verizon as the provider That's it! IP address was set right away and I got online in no time, no pesky VZ Access Manager required. In fact, I'm writing this post with that connection. :)

Life with Ubuntu: day of repair and upgrade

2010-06-11T20:15:00.000-07:00

Did two things today: upgraded to a Dell Latitude E6510 (from a Latitude E5400), and fixed X on a desktop.

The laptop upgrade went pretty straightforward. I installed Jaunty on E6510, then backed up the system on E5400, excluding system specific directories. Then I restored it to E6510. Upon rebooting, GRUB gave me an error 15: File not found. At that point I realized that I should have excluded /boot. Now I've whacked the working menu.lst. The problem was with UUID it turned out. After I booted into Live CD, then fixed menu.lst with the true UUID of the new hard drive (found by ls -l /dev/disk/by-uuid), it booted up fine.

The booted Ubuntu had no GUI though. X said device not found. I fiddled a while with xorg.conf, then decided to just download NVIDIA installer. That fixed the problem in a jiffy. After that it was pretty much smooth sailing. The only problems I had were a corrupt TrueCrypt volume (probably because I was using it when tar did its backup), which I simply re-copied, and VirtualBox complained about not being able to start network, which was fixed by merely opening the network settings tab on VMs and clicking OK.

The desktop computer has Jaunty x64. Its X broke down right about I struggled with Intel Matrix Storage. Not sure if it's related, since the Linux drive was a completely separate one from Windows ones. Anyways, X just errorred out with "unable to initialize PCS database". I didn't deal with it for a while. Today inspired by the success with NVIDIA, I decided to go the same route with this computer's HD 4670s. I downloaded ati-driver-installer-10-5-x86.x86_64.run from AMD. Ran it, and then "sudo aticonfig --initial -f". Restarted, held my breath, and I got my desktop back!

People loved to hate the big 2 GPU companies for their Linux support. But today, I just have to tip my hat.

Pain in the you know where: Intel ICH10 and Matrix Storage

2010-05-15T11:49:00.000-07:00

The SATA RAID supplied by the chipset has proven to be such a pain that I have to vent here. I tried to use it on my Windows 7 Ultimate x64 to provide some redundancy, and eventually gave up after so many failures. The driver seemed to randomly put the array in rebuild, which slows the Core 2 Quad 9550 with 12 GB memory to a crawl. Sometimes a drive was just randomly reported as failed, the whole array as offline. Luckily enough most of the time the boot/system drive was a single Raptor, so Windows still survived. A measure for redundancy for my data turned into a constant live test for my backup strategy. ;)

If you search for Intel Matrix Storage Manager problem, thousands of posts easily turn up. Like this one on Intel's user forum, with 500+ replies. And there are multiple other threads like that, just on Intel's forum. That oughta tell you something... A lot of the complaints are with version 8.9.0.1023, and some reported success with reverting back to 8.8. I had no such luck. All other versions fared just as badly. I also saw reports on 9.x with the same problem. I don't think there's anything wrong with my hardware: once I put the drives on the lowly Windows dynamic disk mirror, things have been just fine.

I've used consumer fake RAIDs from VIA and nVidia in the past and never had much of a problem. Neither was run in a typical environment (VIA for Win XP x64 and NVIDIA for Win 2003 Standard x64), but they were just humming along happily. The nVidia machine even ran VMWare Server with 4 VMs. But the Intel stuff doesn't even work under a lightly used Windows 7 machine.

I couldn't believe Intel had unleashed such buggy stuff on users. Bad vibe for the overall reputable company. They need to get their acts together. If not, I guess there're plenty of chipset vendors to choose from. They may be the king of CPUs, but outside of that, make the steps carefully please!

How to install Salesforce.com Connect for Office with MS Office 2010

2010-05-14T21:00:00.000-07:00

Well, it doesn't by default. The installer runs without error, but there'll be no Salesforce.com ribbon available. I don't know what went wrong, but you get nothing in Office14/xlstart directory. If you do the same thing with Office 2007, you'll get a SFDC.xla, and a SFDC12.xlam in the directory. I tried this on two separate Windows 7 instances, one virtual, one physical, both x64, with the same result. Office 2010 is x86. Connect for Office is 1.8.1.1.

Since the juice is all in the xla files, I decided to just copy the 2 files from a working Office 2007 machine to an Office 2010 VM. Files copied, open Excel 2010, and voila! Salesforce.com ribbon works as expected. I could log on, import reports, refresh them, all without issues.

The source machine for the files has Connect for Office 1.8.0.8.

BTW, it looks like Office13 is skipped. Why do people have to be that superstitious?

90W 9T215 doesn't charge E6400

2010-04-19T21:06:00.000-07:00

That's right. So don't buy those P/N 9T215 generic replacements that claim to be Dell Latitude E6400 compatible. I tested original 9T215 on E6400s (not on purpose) and it just doesn't charge. BIOS says it's unknown adapter. Computer actually runs on A/C but battery doesn't get charged. Why? There must be love between the two for it to charge? It's enough to power the machine, it must be enough to juice up the battery. That's just annoying!

Crippled "Recovery Disc Creation" on HP Mini

2010-04-07T12:27:00.000-07:00

HP Mini 210 is really a good netbook: attractive design, broad configuration options, and very capable offering of software, esp. for cloud and data management. However there's one thing that is quite annoying with it: the "Recovery Disc Creation".

It is cheap stingy for HP to not provide a factory recovery disc (Dell provides that by default). So HP includes this Recovery Manger program (by Cyberlink) for you create a set of restore discs. You'll quickly find the option is really a pain to use, because

Very slow start. Will take more than 10 minutes for it to be ready to burn the first disc, prob. because it needs to decompress source files.
Doesn't take RW discs. Only DVD+/-R is accepted. Why? Did the designer just get teleported from 2003?
The long process will restart if disrupted. If unfortunately the program gets interrupted on the way (computer goes into sleep, for instance), you'll always have to restart the very long process from scratch. No resume option at all.

The annoying program is probably not going to be a deal-breaker, but definitely bad enough for me to write this post to vent... :)

Be thankful this time

2010-03-31T06:52:00.001-07:00

SCO's case against Novell concerning the copyright of Unix related to Linux is finally dead. I never considered it as having any merit, but decisions defying logic could certainly happen in a court. We should all be thankful that is not the case this time.

It has been a test on how well the open-source model can stand against people's urge to keep all profit to themselves. The whole saga should make people think twice before trying to profit at other's expense: just because a certain way of business makes it difficult for everyone to keep profit all to himself, it still doesn't mean everyone would treat the model as an enemy.

I'm not as a staunch cheerleader for FOSS as Steven J. Vaughan-Nichols is, but I'm certainly a strong supporter of it. The whole open-source way happens to bring out the best of us social animals in my opinion - everyone else benefits when you prosper, when doing it right. This makes less useful for our instinct to exploit and fight others to preserve our own genes, because the goal for your own benefit is aligned with others' welfare, and that makes us less animal and more... human, shall we say?

in reference to:

"Novell Wins! SCO Loses!"
- Novell Wins! SCO Loses! - Computerworld Blogs (view on Google Sidewiki)

Migrating Ubuntu ... no sweat

2010-03-08T21:27:00.000-08:00

Here's how I moved a Jaunty installation from one computer to another. The old one is Core 2 Quad Q6600 on LANParty DK X38-T2R. The new one has Core 2 Quad Q9550 running on Gigabyte GA-EP45-UD3P. Different memory and drive configuration. The only thing shared was a pair of HD4670. I just moved the drive containing the root partition to the new computer, booted Live CD, and did the following under grub:
root (hd2,1)
setup (hd2)

Done. Reboot, select the drive to boot from, and I was back in my old Jaunty desktop. Could it BE any easier? It actually could, if I hadn't install grub on another boot drive in the old computer. I wouldn't have even needed the grub reinstall. True plug-n-play.

Could anyone imagine the pain if I were to try moving a Win 7 installation?

By the way, the only thing that caused a bit trouble was the RAID volume (Intel Matrix Storage based) on the new computer. After the grub reinstall, one of the RAID drive got dropped so the volume became degraded. GRUB shouln't touch the drive or anything related to the volume, so what the heck? Luckily the Win7 installation on the volume did survive so I was able to rebuild it, but not without complications. I'll have another post on my frustration on this Intel fake RAID junk.

The annoying "Save error: Unable to perform save on all files" from Force.com IDE

2010-03-05T08:02:00.001-08:00

Sometimes Force.com IDE would throw such generic error when trying to save to the server, the error mentioned is "An unexpected error has occurred". Right, that's helpful... The Log Viewer will have "Save failed!(Open log file for full message and/or stacktrace" and I never found a good way to "open log file" other than checking the debug log from the Web UI (Admin Setup > Monitoring > Debug logs). Mostly I found the error is the result of using field values without including them in a query, or a failed System.assert call. Why couldn't the IDE just say so? There are a host of other things that may cause this too, which makes debugging a pain.

Thunderbird Browser association lost after installing FF 3.6 on Jaunty

2010-02-26T20:30:00.000-08:00

Thunderbird 3 was working fine until the upgrade. Then clicking links in emails just stopped working. I click, nothing happens. Eventually I figured out that the easy way (or the only way) to change the dead browser association is to try the links without FF 3.6 open. When I did that, Thunderbird asked what it should use to open the link, then I got to point it to the right thing. Not sure why, since I verified the default browser setting already, and FF 3.6 was from the Launchpad repository, not a version I downloaded directly from Mozilla...

Live a Ubuntu life: SCIM

2010-02-16T09:03:00.000-08:00

I needed to use SCIM for a number of applications on Jaunty, including Firefox and aMSN. The problem is with apps that weren't installed from Ubuntu repositories. The supplied Firefox 3.0 works fine, but not 3.5 or 3.6 I downloaded from Mozilla. There's no Ubuntu version of aMSN for that matter. The problem lies with QT it seems, which SCIM needs to run with and be compatible with the QT used to compile the target apps.

I solved the aMSN piece with help from this Ubuntu community documentation, but Firefox just wouldn't work. Eventually I installed FF 3.5 and 3.6 from Launchpad repositories, and that brought those apps in line with SCIM. BTW, that also solves my Java plug-in issue with them.

How to change port for NX

2010-02-15T08:55:00.000-08:00

Sometimes you want your SSH daemon to run on a non-default (22) port, and this would affect your NX server as well, as it's based its service on SSH. So besides making the change in SSH, here's the places you need to modify for NX:

/etc/nxserver/node.conf
change SSHDPort to your custom sshd port
This file is /usr/NX/etc/node.cfg for NX Free Edition (non OSS version)
(this seems to only apply to non-OSS version)/usr/NX/etc/server.cfg
change SSHDPort and SSHDAuthPort to your custom sshd port

VBox goes defunct

2010-02-02T07:04:00.000-08:00

Sometimes when I use script to start a VirtualBox VM - usually I use that if the networking option needs to change by location - the VirtualBox process goes defunct. Then no subsequent run will go because of that stuck process. I'd have to find its parent (VBoxSvc) and do a "kill -9" on it.

Dell resells Salesforce.com now!

2010-01-22T14:35:00.000-08:00

I was approached by our Dell rep about whether we use CRM and if we'd interested seeing "what they use". Seeing the good buddies Mike and Marc on stage at Dreamforce, I sorta knew what's coming, but a little in disbelief. Sure enough, the rep claimed that they are the only reseller right now. I was like "wow!" (silently), for now SFDC is in its Walmart phase - it's being carried in a virtual Walmart for IT stuff like Dell. Guess you can't get any more main-stream than that.

After some googling, it turns out that the deal was announced late 2009. Can't believe I missed that news... Three cheers for cloud computing!

Easy as pie - Flash plugin for Firefox on Ubuntu

2009-12-21T08:27:00.000-08:00

It already is, if you just want it on the standard Firefox on the system. But what I needed was to have it on FF3.5 on Jaunty, which doesn't come by default. I have FF3.5 installed in my home directory. Turned out all I needed was to download the .tar.gz from Adobe, and put the .so file inside to the plugins directory. I was trying to symlink from the standard /usr/lib location to no avail.

apt-cacher bug in Jaunty

2009-12-01T08:54:00.000-08:00

I ran into this one. Basically the apt-cacher in the Jaunty repository (1.6.4ubuntu4) throws some error messages - "Error reading from server - read (104 Connection reset by peer)" - for some item checks requested by clients. It's fixed in upstream Debian Squeeze (ver 1.6.9) as far as I can tell. I had to download .deb directly from Debian, then used "apt-get -f install" to fix some dependency issues to dpkg it.

Another interesting issue I found is that the parser for /etc/apt/apt.conf.d/01proxy file in apt-get apparently ignores the proxy address, if you forget a semicolon in the header - I did precisely that, so having Acquire::http:Proxy in there - then it just silently went out without using proxy. It drove me nuts since there's no error message, until I paid close attention to my typing. :)

UPDATE: Now I ran into another bug, which prevents me from upgrading machines to Lucid. So I had to remove Debian apt-cacher (even their newest 1.6.11 doesn't work), and put on 1.6.7ubuntu4, which still has the original bug. Lesser of the two evils. Why can't Ubuntu and Debian address both?

VBox audio driver for Windows 7 x64

2009-11-20T12:47:00.000-08:00

I downloaded it from here. It's a Realtek chip. Please note the driver is unsigned (not recognized by Windows 7).

Live a Ubuntu life: burning DVDs

2009-11-18T08:21:00.000-08:00

Today is the first time I had the need. I have a Win7 installation ISO I need to burn on to a physical disc. I right clicked the file and selected "Open with Disc Burner". That opened up Brasero, then I realized I didn't actually have a burner on the system.

No worries. I plugged into eSATA a Dell E-module I could find, and put in a DVD-R blank. The disc was spinning for a long time, and Brasero still said "no disc available". So I took a closer look at the drive - it's a DVD-ROM. What a dummy! Who keeps DVD-ROM drives around anymore these days?

I went looking again and found a Buffalo external Blu-ray burner. That oughta do it. Plugged in the power and USB, inserted the blank disc, and in a minute Brasero saw the disc and I was able to start burning. A few minutes later I got my Win7 install disc.

So, except for my own dumb act, it was an entirely smooth operation. A few years back, Linux desktop distros were still deemed too troublesome to use (by some I should say). Now it challenges the best experience Windows or OS X can offer, with only a fraction of the hardware vendor cooperation those guys got, not to mention billions of dollars wasted on copy right protection and marketing. That's what I call bang for the buck.

"Gone Google"?

2009-11-18T08:08:00.001-08:00

Saw this ads by Google on Network World:
"Over 60% of Fortune 1000 have gone Google".

Of course there's a fine print attached to such claims: "Gone Google means using one of the Google enterprise offerings, from Google Apps to Postini...".

Sounds like a typical way of making one sound more important than actually is. Using one of your stuff means they are now in your camp? Come on. In fact, if there's one company doesn't have to do that kind of cheap marketing hype, it's Google. Doing that is just ... cheap.

On the other hand, the true target audience may not be as dismissive as I am. There are surely going to be people who start to sweat on the claim. M$ anyone?

Live a Ubuntu life: KeePass on Mono

2009-11-17T18:50:00.000-08:00

I used pcorral for password safe on Windows. It's a nifty little program that allows me to carry around my huge list of passwords safely on a USB stick. When moving to Ubuntu, I wanted to do even better - something that allows me to do it cross-platform. Eventually I picked KeePass, which is a .NET program, so it can be run easily on Windows and Linux.

KeePass requires Mono 2.2 or higher on Linux. Since Jaunty only has 2.0.1 (Karmic is on 2.4), I had to enable the PPA source (monoxide) to install the newer Mono build. After that's done, KeePass was up and running in a minute. Cool!

Live a Ubuntu life: VirtualBox

2009-11-10T05:36:00.000-08:00

I've been using VMWare Workstation on Windows. USB support is the main reason I stayed with it instead of going for free VirtualPC. Now I'm on Ubuntu, I want to try the free option again. This time it's Sun's VirtualBox. I didn't go for the open source one, as Sun has put some serious limitations on it. With the free-to-use edition, things are looking really good.

I had no problem setting up Windows 7. I allocated 2 GB for the VM and 40 MB video memory. The only issues I had with running it were sound not working, and I couldn't use VirtualBox video driver, which results in limited choice for screen resolutions. Neither is a big deal, but I'll try to figure it out.

Migrating a Windows XP virtual machine from VMWare worked even better than I expected. I followed the two advices here, then just told VirtualBox to use the vmdk's. The VM just booted and ran like a champ! I actually forgot to do mergeIDE before migration, but I may have done it at some point before since the VM has survived previous migrations.

One sticking issue is that PPTP VPN doesn't seem to work on a NAT'ed guest for now, which I hope gets fixed soon.

UPDATE: I was able to re-enable the virtualbox video driver in Windows 7. Not sure why the first time after installing it didn't work.

Live a Ubuntu life: application scorecard

2009-11-09T08:56:00.000-08:00

This should be a good way to track experience with a new computing environment - whether the things you need to do are now easier and harder to do, and whether they are done better or worse.

Windows app	Linux app	Experience grade	Notes
Pidgin	Pidgin	better	Like the integration with notification applet; no separate download
Remote desktop	Terminal services client	same
vnc viewer	remote desktop viewer	same	VNC into some Windows server seems to be a bit lagging in cursor movement
VMWare Workstation 6.5	Sun VirtualBox 3	same	price is definitely better!
Mapped drive	Connect to server: Windows share	same	CIFS server/share name work fine with no manual WINS config; hidden shares work fine

Live a Ubuntu life: dual screen setup

2009-11-09T06:38:00.001-08:00

I just have to have it. In my previous Windows life, I had no problem setting up Windows XP and Vista with a NEC MultiSync 90GX2, and a BenQ FP202W. I don't want to buy new stuff just to get the same amount of screen real estate.

It was a little harder than I expected. Took some command line rescues and online searching. Eventually the config working for me is like this. Since the graphics core in my laptop (Latitude E6400) is Nvidia Quatro NVS, the tool to use is nvidia-settings coming with the Nvidia binary driver. I used separated X screens with Xinerama enable. Both screens have auto resolution. That gives me the ability to freely move windows across the two monitors. I can still use multiple workspaces too.

I like my Compiz eye candies, but there's no way to get them working. Some suggested using twinview with no Xinerama. It didn't work for me. Well, it did, only for a minute or so after logging in. Compiz effects worked, but then the whole computer would lock up solid, requiring power button treatment.

If I use the laptop by itself, Nvidia correctly falls back to one screen setup on LCD. And all Compiz stuff will be alive again, which is a nice bonus.

I've yet to find a way to use 3 screen setup, with laptop LCD in the mix. But it should definitely be possible.

Live a Ubuntu life

2009-11-09T06:21:00.000-08:00

The title is a bit bigger than what I actually mean, but I like it. After all, "ubuntu" means humanity to others, which I definitely set as my life's benchmark. Philosophical discussion aside, this should start a series of postings about my transition into a Ubuntu based computer. It's a fairly drastic change, considering I work in and manage a mostly Windows environment.

So the first question is probably "why?". Well, I want to. I'm not a Microsoft hater, or an open-source purist. I am very practical. I hold multiple MCP certifications, but use open-source whenever applicable. I like the model of open-source, because I think humanity is at its best when we're in a collaborative mode. It's not always easy to make it sustainably workable, but like many good things, you just have to work for it. Also from a tech geek's point of view, using a proprietary only toolset greatly impairs one's ability to learn in-depth. After all a fundamentally blackbox setup is not encouraging for under-the-hood exploration.

A smaller but related "why?" is "why Ubuntu?" Simple: I like it the best. I've tried at least five other well-known distros. This Debian based gem combines the depth and ease of use the best - again, like humanity, the ones you like the best will probably be those with depth while being approachable.

With the "why" questions out of the way, a long series of "how" questions will follow...

A heroic way to die... for a computer

2009-10-29T19:43:00.000-07:00

If you have a lengthy enough history with computers, you probably saw a computer's hard drive die at least once. On a Windows computer, you'll most likely see a BSOD, then upon restarting, your computer tells you that it doesn't think your hard drive exists.

So what would a Linux computer do when a similar disaster strikes? Well, you probably wouldn't know for a while unless the computer is actively monitored. That's what happened to me. I had a laptop computer running Ubuntu 6.06 (yes the old trusted Dapper) Server. The first sign of trouble was that DHCP clients were no longer getting IP addresses from it, which served as a DHCP3 server. It took me a few moments to figure out what was going on. The tell-tale sign was a console message like this "Journal writing failed... Abort... root remounted read-only". Well, that's the machine's way of telling me "well, I'm not seeing the disk anymore, so I'm just going to treat it as read-only and move on". The machine was still running, and it still did some other duties fine (like routing stuff between two subnets). But many commands no longer ran with "I/O error", including the shutdown. Once I restarted it with the power button, I saw the familiar "primary disk not found" message.

I was thoroughly impressed. Imaging a warrior having his legs cut off and even his ground yanked away from under him, but he keeps on fighting? That's what the O/S did. Maybe I'm naive - hey, I was Windows only one time. :)

BTW, that Dapper was running on LVM.

MiniNote 2133: Reloaded

2009-10-25T20:39:00.000-07:00

These days virtually every netbook is powered by Atom. However I'm not willing to let go my HP MiniNote 2133 yet. Since I can't swap out its C7M heart, I'm going to beef up other parts. Mine is Model KR922UT, which is the lowest config, so there's plenty of room to go up. First I go max on memory with 2 GB of PC2-5300. Then I put in an OCZ 64 GB SSD (OCZSSD2-1SLD60G) to replace the default 4 GB SSD. The first change was very straight forward, the second was a little more than that. The 4 GB "SSD" from HP really looks more like a piece of circuit board than a drive. As such it's smaller and thinner than a standard SATA 2.5" drive, and was installed within a metal bracket. There's no way for me to fit the OCZ drive into the bracket, so I just ditch the bracket. The OCZ drive fits quite well in the space, but not as snuggly as I'd like. If you shake the netbook, you can feel the drive moving a bit inside. Not that SSD drive would be harmed by that level of looseness, but I prefer things to be a tad tigher than that. My solution? Two pieces of thick double-sided sticky pads. After that, everything is firmly in place.

Once the hardware is refreshed, I proceeded to revamp the system software as well. As much as I was pleasantly surprised how well SUSE worked on the anemic configuration, I'm ready to use my beloved Ubuntu again. I tried to install Xubuntu 8.04.1 first, thinking that might go easy on the little machine, but the GUI installer wouldn't load. Not wanting to tinker with it more, I tried a regular Jaunty installation. It went through flawlessly, and everything I care works fine. It's fast too, taking only 62 seconds to boot to the desktop, and 75 seconds to start using Firefox. Not too shabby for an aging C7M powered machine.

BTW, I added berry4all and used it with my Verizon 8330. That worked fine too. I found out that I shouldn't disable wireless from Network Manager, as that would break how pppd write to resolv.conf, which leads to name resolution problem.

Overall I'm very happy with the results. Armed with this rejuvenated little netbook, I'm ready to roam! Just can't ask for a better outcome for a $160 upgrade.

Dub WebEx recording with .wav files

2009-09-08T14:03:00.001-07:00

That's exactly what I needed to do, but the menu option for that in WebEx Recording Editor 2.8 gave me "invalid wav format". The suggestions in the discussion following the blog post were very helpful. Basically the wav file has to be in CCITU u-law format. The Sound Recorder in Windows XP can make the conversion very easily. Not the one in Vista though, which doesn't seem to allow opening and manipulating an existing wav file.

The tech support from WebEx really shouldn't just say "it's not supported", which probably was the easiest to do. Give the most details they could find and let the users hack their way. Again, it's a rare thing to find a tech support with spirit these days.

SFDC Connect for Office: Better together?

2009-09-01T09:35:00.001-07:00

del.icio.us Tags: Salesforce.com ConnectforOffice Excel Microsoft

It really depends I guess. It's a great tool to bring a Salesforce.com report into Excel, which has a rich set of presentation and data processing features. However with slim documentation on how the software works, a few caveats really should be noted for any serious uses.

Basically Connect for Office works by using Excel's Data Connections feature. However you can't edit any connection created by it. This is understandable for reducing the chance of user messing up the inner workings, but also limits your options. It'd be great to be able to replace an imported report that way - all I need to do is just changing the URL of the report source - then I get to keep all the formatted calculations and charts based on the report. Well, you know life is never that kind, is it? For that kind of change to happen, you'll have to go through the tedious steps of importing the new report, copying the dependent elements (formatting, formula, charts, etc.), removing the old setup, and adjusting anything that might break.

Please note the order of those steps for altering source report is the best practice to follow. Any deviation from that will be even harder to do (say removing the old report first) due to the fact that there're hard coded behavior - some from Excel, some from Connect for Office - that can't be changed: imported report only inserts (never overwrites), Excel adjusts references in an unmovable "smart" way, etc.

And don't forget this: never change the worksheet name after you imported the reports, because the names are the IDs for the data connections to work. It may be Connect's developer's fault, but could also be the result of Excel's API. I have no interest to find out who's to blame, but this makes life very hard if you have to change the sheet names. So try to be generic: for instance, anything says FY 2009 will get you into trouble if this is going to be reused next year - you just have to redo things all over again.

Flash 10 kinks with FF 3.5

2009-08-31T09:08:00.000-07:00

Sometimes it's a matter of figuring out "what have I just done?".

All of a sudden my URLs in emails stopped working. Outlook will say "cannot find the file specified", and Thunderbird just does nothing when clicked. I toggled default browser, and IE 8 would work fine, but Firefox 3.5 still no go. I found a report on Sophos scanning plug-in causing the issue. Of course I don't use it, but it did remind me that I just installed Flash Player 10 to check out a friend's new pictures from Tibet on photoshop.com. That action added an extension (content download manager) as well as a plug-in (Flash 10.0.32.18)

So I removed the extension. No help. Then I disabled the plug-in. URLs came back to life! Then I re-enabled the plug-in, and things stayed OK.

Something not very smooth between Flash 10 and FF 3.5.2 then. As usual, they need to do a better job testing. Granted I'm an advanced user with unique needs sometimes, but look at this long thread with annoyed users and webmasters alike, you know they could definitely do better...

VMWare Server 2 & Ubuntu kernel upgrade

2009-08-21T20:26:00.001-07:00

Today is the first time I ran "sudo apt-get dist-upgrade" on the machine that runs VMWare Server 2 on Hardy Server. It's actually a lot painless than I expected. All I needed to do was to run "sudo vmware-config.pl" after the kernel upgrade. The only thing I didn't quite know to do was to have the headers ready before the vmware recompile. That's just a "sudo apt-get install linux-headers-$(uname -r)".

Then I sorta did the same thing on a Ubuntu guest (Hardy as well with a generic kernel), expecting to run vmware-config-tools.pl. But that turned out to be unnecessary. Not sure why, as I had so much trouble last time.

Fixing a non-starting firewall on Vista

2009-08-19T19:00:00.000-07:00

Windows Security Center informed me that WFW isn't running, but it couldn't start it. Event log shows ID 7024 with error code 0x5. Googling around led me to this thread, which towards the end suggested KB 943996. Basically the KB suggested to check permissions for account "NT Service\MpsSvc" on several registry keys. This first key I checked and corrected was
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\SharedAccess\Epoch (needs Query Key and Set Value permission),
and then the Firewall started!

How did that permission got lost?

By the way, still don't know what is the NT Service\MpsSvc account. You have to search that exact name to turn it up on the permission setup screen.

Encode VOBs into xvid AVIs on Etch

2009-08-06T08:30:00.001-07:00

The output would mainly be consumed on Windows client. The workhorse I can use is a virtual Debian Etch on XenServer 5.0. I tried the ffmpeg installation from Etch native repository. It worked fine but it can only use FMP4 for default video codec. Windows Media Player / MPC just didn't play the video on my Vista x64 with Klite installed. I was searching for a solution. Some suggested configure ffdshow to support FMP4, which it already did ("other MP4" enabled). Some suggested changing the 4CC of the file, which looks like too much work. As I was fudging along, I wanted to double check what error message WMP had, if any. But this time it played! Weird. All I did was checking the ffdshow config and using G-Spot on the file...

Anyway, this is all too unreliable. I gotta produce xvid AVIs. So I chose to use debian-multimedia.org. Pretty easy to use. After I added the repo to sources.list, updated, ffmpeg is ready with xvid. The syntax that worked for me is just

ffmpeg -y -i VTS_01_1.VOB -vcodec xvid -b 1000000 -f avi my.avi

For multiple VOBs, it'd be a simple concatenation:

cat VTS_01_0.VOB VTS_01_1.VOB VTS_01_2.VOB | ffmpeg -y -i - -vcodec xvid -b 1000000 -f avi my.avi

BT file transfer from BlackBerry 8330 to Dell Latitude D630

2009-08-04T19:27:00.001-07:00

I largely followed the steps detailed here. Once the two paired up, I had two issues. First, BB device software promptly crashed, rebooting itself. Second, Vista is asking for driver for a new device. The second issue seems to be OK according to people here - basically just tell Vista "don't bother". The first one might be tougher to solve... (to be continued)

Unscientific comparison of Vista and SUSE startup speed

2009-08-03T13:17:00.001-07:00

It just so happened that I started two machines almost at the same time, and side by side. The SUSE machine is an original HP MiniNote, with a lowly VIA C7, 512 MB memory, and 4 GB SSD. The Vista machine is a relatively beefy Dell Latitude XT2 with Core 2 ULV 1.4 GHz, 3 GB memory, and 80 GB 5400 RPM HD. Neither has much additional software on beyond the factory default. I didn't time the two, but since realized the interesting comparison, so I did note the time difference. From power on to browser open (Firefox for SUSE, IE 8 for Vista), MiniNote was only about 10 seconds behind XT2. Both startups were between 2 to 3 minutes I think. I'm sure SUSE 10 was optimized for the setup, while Dell's Vista install probably less so, but still the slowness of Vista was too apparent, considering how well equipped that XT2 is.

Thunderbird not able to deal with attachment that has special characters in name

2009-06-24T12:31:00.001-07:00

I'm using the Portable Apps edition of Thunderbird (2.0.00.22). The attachment has a few %20 in its name because it was downloaded. When trying to send the message, T-bird would say "can't open .... Please check "Temporary Directory" setting". After removing the %, it worked fine.

Jaunty on HD 4670

2009-06-24T09:43:00.000-07:00

The Vista drive died on my desktop so I decided to install Ubuntu Jaunty x64 instead. The computer has a pair of HD 4670 in CrossFire. I had to go through safe video to install it, otherwise the screen just blanks on me.

X seems only recognising resolution up to 1280x1024. It's not detecting/using the 1920x1200 on the KDS 24" monitor. I fudged with xrandr and xorg.conf but couldn't get anything to work better. Then I moved on to try fglrx (open source ati driver doesn't support 4670 yet). The one in the repository made X fail, since fglrx.ko isn't at the place it's supposed to be. Don't know if it's a x64 issue or not.

Then I downloaded installer from ATI, not hoping for much, but it works like a miracle! Installer went through fine. Log off, log on, voila, I got 1920x1200! I never really was fond of ATI's drivers, but this one did shine.

Tried a H.264 1080p video under Movie Player (Totem) and it ran fine. Action sequences were smooth, but it would slow down visibly if the control menu is shown. Probably a player bug. In fact if the controls are gone and I move the mouse to get them, the screen would go back to desktop background briefly before returning to video with control bar. Annoying.

SATA controller in E6400

2009-06-22T11:13:00.001-07:00

It's almost unworkable with Windows XP. To get it to be useable for XP, I'll have to set it to IDE mode in BIOS; AHCI and IRRT won't work. There're a few links relevant on getting those modes to work with XP, but so far I haven't figured it out how exactly in my case. There're just a lot of moving parts.

http://en.community.dell.com/forums/p/19264477/19450243.aspx

http://blogs.technet.com/benhunter/archive/2007/07/12/bdd-2007-integrating-mass-storage-drivers-into-your-xp-image.aspx#1563162

In addition, Acronis Disk Director and GPartEd seem to stumble too on the controller, even with IDE mode. Acronis would say no disk found, and GPartEd would be stuck with I/O error upon booting.

Acronis 2009 Disk Director Home doesn't support Latitude E6400

2009-06-22T08:35:00.001-07:00

I've already set the SATA controller to operate in IDE mode (otherwise TrueImage won't recognize it either), but Disk Director still says no hard drive found.

BTW, TrueImage 2009 is really fast in generating image. A 10 GB partition of Windows XP only took 10 minutes to be imaged to a network drive.

Mounted volume issue

2009-06-12T07:54:00.001-07:00

One of the mounted volumes got moved to a directory (prob. by an accidental drag-n-drop - need to test the behavior though), so I tried to recreate the folder and mount the volume to its original location. I got "folder does not exist or already a drive path to some other volume". WTH?

Some googling turned up this link http://groups.google.com/group/microsoft.public.windows.server.general/browse_thread/thread/570ac48bce7413ba?pli=1
Essentially you need SYSTEM to have full control on the empty folder to create the junction point, but the created folder inherited the parent, which was configured to not have that - probably an oversight when setting up sharing.

Dell Latitude D630 not able to support Blu-ray playback

2009-05-26T13:14:00.001-07:00

I ran PowerDVD 9 on my Dell Latitude D630, which has Core2Duo T7250, 4 GB DDR2, and Nvdia NVS 135M, running Vista x64. I was surprised how bad the quality was with Blu-ray playback. There was a significant amount of frame loss. I used to think the laptop as being able to do some heavy lifting, but apparently not so, at least for Blu-ray decoding. I'd guess the bottleneck is with the graphics core (rated at 3.8 by WEI, lowest among all), which isn't designed for this kind of task.

UPDATE: Same software works fine on a Dell Optiplex 760 with Core2Duo E7300, 2 GB DDR2, Radeon HD 3450, and Vista x86. GPU score is 4.3. CPU might be a factor too, now I think. The WEI on this machine is 5.5, while on the Latitude is a meager 4.9.

Dell 968 refused to install program on a Pentium M 1.7 GHz laptop

2009-05-11T14:15:00.001-07:00

It said processor speed was too low. CPU was running at 347 MHz (according to Computer Properties), probably the result of SpeedStep. Can't believe the installer could be that clueless about CPU speed... Anyway, went into Power Options and disabled CPU power management. After restarting the installer went ahead. Will turn back on the power management later.

Windows 7 RC (x86) on VMWare Server 2

2009-05-11T13:08:00.001-07:00

The installation was mostly eventless - and how cool was that! VMWare runs on Ubuntu Hardy server and I selected Vista as guest type (adjust memory to 512 MB). The only issue was at the step of setting password for the first user. The O/S froze up for a few minutes. The mouse cursor was barely movable... I could tell something was hogging CPU, since VMWare portal showed full usage of the CPU. Things were fine after that, except that the next few config screens (time zone, etc.) just flew by, probably because I pressed Enter too many times during the freeze. :)

Windows 7 feels pretty snappy, even running on a modest P4 2.8 GHz with no virtualization assist and 512 MB memory.

Linksys QuickVPN

2009-04-30T18:21:00.001-07:00

QuickVPN seems to be a pre-packaged VPN setup that's between Linksys routers and PC clients. Not sure if it's IPSec or SSL - it does use 443 (or 60443) though. Anyway, because its canned setup nature, there's no room for any interoperability with any other VPN platform.

Client log issue for NX

2009-04-29T17:53:00.000-07:00

I'm running a NX node on a Ubuntu 7.04. Usually I'd log on to it from a Windows client on Vista x64. Lately the sessions seemed to die after a period of time. Then I noticed sometimes it popped up messages like these:

Found some clues from this page, even it's in a language I don't know (some sort of Nordic?). So I disabled the error logging of X, and doubled the log file size limit to about 8 MB. Hopefully this would give me more lifetime of my sessions.

The NX included by Ubuntu seems to be old (2.x). I wonder maybe it's dropped by the distro?

Battlefield: Netbook

2009-03-31T20:53:00.000-07:00

Windows 7's mission: reclaim the netbook territory.

If anyone still has doubt, check out the Technet homepage. The second featured article on Windows 7 is about using USB drive to install Windows 7 on a netbook, which by the way has been common for Linux for quite some time.

Netbooks may well be place for the beginning of the end of Windows' dominance in the client OS market. With Android on netbooks around the corner, Microsoft clearly sees the writing on the wall. There's really no time to waste for them, and it'll certainly be very interesting to see the battle.

.NET 3.5 SP1 as high priority patch by Win Update - bad & bad

2009-03-04T12:27:00.001-08:00

I don't understand why. Computers that don't need .NET at all are getting hit with that. It just feels like a blatant effort to push the .NET platform. And holy cow, it's a D/L bigger than Delaware! Even on a 6 Mbps line, this 287 MB behemoth could take you a good 10 minutes. I wonder why there isn't a bigger outcry on this. Only found complaint from here:
http://www.askwoody.com/2009/2284/

Drives for Dell EMC AX150i

2009-03-02T11:53:00.001-08:00

In my continued effort to tinker with AX150i, I tried a WD5000AACS in an EMC tray. No go. The array say it's not supported. The alert message only says a SATAII is needed. The drive is one of the Caviar Green series from Western Digital. 500 GB 3gb/s. I don't think it's much different from the Barracuda ES 250 GB I got from Dell/EMC. I'm not sure if it's NCQ though. Wondering how the array determines compatibility... itching to reach the conclusion of vendor lock-in conspiracy, but willing to try some others before we go there. :)

UPDATE: I tried a Barracuda ES 500 GB. No dice. Now the conspiracy is completed. They indeed want to lock you in with their expensive drives; or they really don't want you to have an easy and economic way of scaling. Either way, coupled with lack luster support I got so far from them, EMC just slipped to my trash can end of my vendor list.

GPartEd 0.3.9-4 crashed

2009-02-27T20:13:00.000-08:00

I never had problem with GPartEd before, even it's not 1.0 yet. So this surprised me. It happened when I was resizing a cloned partition to fill a bigger drive. Two partitions: 1 NTFS, 1 FAT (IBM diagnostic partition). So I moved the FAT to the end of the drive and expanded the NTFS. Once I clicked "Apply", the GPartEd window went away. I kinda scratched my head and opened a terminal window. top showed gpartedbin is hogging about 97% of CPU. I waited for a while. Nothing. Guess it's stuck. I tried opening another instance of gpartedbin, and it showed partitions unchanged but NTFS having unreadable content. Crap! I had no choice but to kill the thing. Luckily things seem to be OK after restart.

Eventually I got the job done with Acronis. Not sure if this is a bug.

The dirty little secret of Dell EMC AX150i

2009-02-26T09:50:00.000-08:00

I got it really cheap - around $4000 with 3 250GB SATA drives installed. It seemed to be a very good deal considering its functionalities. But when I tried to extend its life and upgrade it a little bit, I was rudely surprised. Dell quoted a 500 GB module at more than $600. SATA drive at that price these days? All they provide more than a generic drive is just a tray so it fits the array. Are you kidding me? The 750 GB module is even more obscene at more than $900. Even when their site is selling the 250 GB module at $69.99, they couldn't explain why it's so much more expensive with the bigger drives. The array model isn't that old!

Dell also quoted a Next Business Day support contract at more than $2,200 a year. They seem to entirely rely on EMC guys to quote that, even the EMC support guys in the past told me to talk to Dell for any support issue. At that price you can buy a new, better array for the cost of a two-year support contract. Yep, that's how they promote their new wares.

It's probably true for all the so-called low cost iSCSI arrays. The "hidden" cost of upgrade and support makes them not so good as the first look. But hey, isn't that often true for many IT investments?

Editing auto-start setting for VMWare Server 2 guests

2009-02-19T14:26:00.001-08:00

It's not that tied to individual but rather a host wide setting list. That's not very intuitive but kinda makes sense. You go to "Edit Virtual Machine Startup/Shutdown Settings" command. Also you use "Move up" "Move down" to shift guest among 3 lists (auto with order, auto with no order, manual) to set their config, which isn't very intuitive either.

Resolution issue with Windows Photo Gallery Slideshow

2008-12-11T09:23:00.000-08:00

This probably applies to both Vista's built-in viewer and Windows Live one. The slideshow viewer would lower the resolution if it thinks your computer can't handle the details smoothly. There are plenty of ways for it to think wrong, and it did in my case. My setup: Vista x64 on Q6600/2GB with Radeon HD 2400/256 MB, 1920x1200. Graphics score: 3.9. The viewer apparently thinks it's too low. So it dropped the resolution to 1280 x 800 I think.

There's a way to force it to use the screen resolution and it worked for me fine. Playing was smooth and nothing blew up. :)

See the tip here:
http://chris-russell-v2.spaces.live.com/blog/cns!6318A7EEC8127EBA!406.entry
or KB930102. Basically it's setting this DWORD value HKEY_CURRENT_USER\SOFTWARE\Microsoft\Windows Photo Gallery\SlideShow\WinSATScore = 500.

First impression with Mini-note 2133

2008-11-11T11:22:00.000-08:00

I got it mostly because PCMall had a sale on it for $304. I have been wanting to try a sub notebook for some time and this is a good chance. If it works out, I'd have a good device to carry around for testing network and doing quick Internet stuff.

The machine came in an unassuming HP box. Only a few things inside: the mini-note itself, battery, power cord, a quick start booklet, and User Guide CD. Install the battery, plug in the power, and turn it on, no manual required. Upon booting a standard SLED configuration screen appears. If you had experience with SUSE before it'd be piece of cake. If not, a little background with Linux should get you by just fine. Even just some general knowledge about computers should be enough to get through too. Then the computer rebooted and I was presented with a proper login screen.

Once logged in, things were pretty much smooth from that point on. I got hooked up with my WPA2-PSK wireless network in no time - well, almost. I went for the Network icon in the Computer menu and got nowhere. Then I realized I should just use the NetworkManager applet on the panel directly - I should know better, as that's I always have done with Ubuntu. The included Firefox 2 was fast and easy. Realplayer played media fine, including online streams. The speakers are of decent quality and power. Evolution was a snap to set up with my Exchange, after I figured out the correct form of URL for the asked OWA server: https://servername/Exchange. Hibernation worked flawlessly, which used to be a pain with DIY Linux installations. The battery life is OK at about 2 hrs. The only thing didn't work correctly the first time was the screen resolution. It started with 800x600. I had to set it to 1200x800 in Preferences and reboot to get it to stay in 1200x768 (not sure why it didn't like the 800 height).

Please note that I had the most basic model, KR922UT. That means a VIA 1.2 gHz C7M ULV, 512 MB DDR2-667, and 4 GB flash drive. Pretty humble stuff but everything seemed responsive and smooth. I did had trouble playing a 700kbps Real video clip, but then I realized it's the SD card it's on. Once I've switched it to a Sandisk Extreme III, it ran like a champion. And there's plenty of room to upgrade: $30 get you a 2 GB memory, maybe $100+ for a 7200rpm SATA 2 drive (or SSD), and there's extended 6 cell battery.

It's really a nice Linux based subnote. You could certainly get it with Windows XP or Vista, but I doubt it'd run better. This is actually the first pre-installed Linux laptop I had (plenty of DIY of course), and I'm very satisfied. Only if they'd have Ubuntu. :) I just feel Windows wouldn't stand a chance if not for its existing monopoly.

By the way, now Amazon has it for $299 and free shipping. Nice.

The way Debian manages new network interface

2008-11-06T14:14:00.000-08:00

When I upgraded my VM Workstation 6.0.4 to 6.5.0, the network for my Hardy Heron guest stopped working. First I thought it's just outdated VMWare Tools. But after I installed the new one, still no luck. I even tried this fix to make sure VMWare Tools does work - basically using the more compatible open-vmtools. No go. Still the same "no such device" message. Then I remembered the problem when I upgraded Ubuntu. It's the system using the new interface as "eth1". Indeed ifconfig -a showed a eth1. So I resorted to editing /etc/udev/rules.d/70-persistent-net.rules. After correcting that, I had to rmmod vmxnet and modprobe it to get eth0 going.

BTW, now I know why Ubuntu ditched its /etc/iftab and went with this long ugly rules thingy - that's the orthodox Debian way. :-)

UPDATE: I also had to manually edit the vmx file to set up a static MAC, otherwise VMWare just divvy up a different MAC every time and the whole thing just keeps breaking.

Dreamforce '08, Monday night

2008-11-03T22:04:00.000-08:00

Man, do they know how to party! A gala with open bar, Foo Fighters, and dim sum just can't be too shabby. I actually stayed close to the back of one of the open bars, which really gave me easy access to beers and a place to sit (on the table). It's also a good spot to watch the party goers. Nerds party too, don't they? To be fair, probably fewer than half were true hardcore geeks, but heck, it's fun to stereotype, isn't it? :) I was surprised to see quite a few young female attendees around, proving the point that more than just IT folks were there. Sorry for being stereotypical here, but there really are more females in other fields than in IT, much to the grief (or pride) of male IT nerds. I did try to talk to one of the seemingly alone women but she said she's not by herself. Good luck to her. I hope she's lying only because I looked like a menace or bore. There were also quite a few guys looking lonely but they didn't seem nearly as interesting. ;)

Eventually I got bored, right at the moment that the lady in front of started to halfly take off her jacket and show her scant tanktop underneath. I decide to walk back to hotel. The night was beautiful, mild and clear after a day of raining. San Francisco is also a non-stop city, maybe not as much as NYC, but definitelly vibrant enough, at least to a hillbilly coming from Detroit suburb. I almost forgot how much I'm excited by the city rhythm. Sometimes I even laughed my dear wife's excitement about a metropolitan life. But when I was surrounded by it, I surrendered.

A few dozen people lining up outside of a Sony store to wait for the release of Resistance 2. Heck, I wish I had passion like that. I don't even have Resistance: Fall of Man yet, even I always wanted to. Power to the determined!

I wish I were at home... seriously. Don't need these mixed thoughts right now. But it's probably just a result of too much booze. Suffice to say, this night's drinking covered a lot of categories of alcholic substances. :-)

Setting up custom networks in Hardy

2008-10-21T11:56:00.000-07:00

If you use Hardy in a regular laptop fashion, i.e., having it automatically connect to available networks, wired or wireless, it works great. Just set roaming mode on both of your adapters and network-manager will take care of the rest. My problem was wanting to have it act a little differently. I want the wired connection to have a static IP, and having the wireless connected to another network at the same time by DHCP. Here's what I found out to be the simplest way:
1. edit /etc/network/interfaces to have this in there:
auto eth0
iface eth0 inet static
2. uncheck "roaming mode" from the wired connection in Administration > Network, and set the static IP info.

Editing interfaces is necessary; otherwise the system for some reason doesn't know how to take care of eth0, which will be dropped by network-manager after taking off "roaming mode" and set static. Keeping "roaming mode" with wireless connection gives you the option to easily pick an available wireless network from the network-manager applet.

"Certificate contains the same serial number as another certificate" - Firefox with WRT54GL

2008-10-21T11:53:00.000-07:00

http://support.mozilla.com/en-US/kb/Certificate+contains+the+same+serial+number+as+another+certificate
You might be hit by this if you access Linksys WRT54GL settings page via SSL - I was anyways. Apparently WRT54GL is one of the routers mentioned in the Mozilla support page that "improperly regenerate their security certificates instead of keeping the same one for the life of the device". Nice job, Cisco!

UPDATE: As time goes by, I started to notice that this happens quite often with self-signed certs, and sometimes deleting the certs themselves doesn't even fix it. The important thing is to also delete the associated CA under Authorities tab, as detailed here by sky76.

SONY DLT IV tapes suck!

2008-09-24T13:22:00.000-07:00

The 3 I bought recently (less than 1 year) have all gone belly up. Those are the newest ones I had. Usually Sony ones are cheaper than say Fuji or HP. Now I know why...

Notes on Hardy

2008-09-19T12:27:00.000-07:00

Did I already say I love Hardy Heron? No matter, I'll say it again any time. Recently I did 3 upgrades to Hardy, 2 from Dapper, completely painless. One from Breezy, which I had some hard time. But that one mostly is the result of running inside VMWare and the way it was set up (I downloaded as a virtual appliance).

The Hardy running on my old Latitude X300 is great. I finally got the old Intel Wifi chip completely working. Well I shouldn't say "I got", because Hardy just took care of it. It found my network, after entering the key and set up the MAC on my router, it just worked. Great work on Ubuntu's part, for they had to retrofit all those old chipsets out there. I also selected some AA font so it looks great on the LCD. Battery life is very good - with a very old extended battery I have, I got almost 5 hour running time with it, on Wifi.

PPTP VPN is one thing that has a tiny issue. Hardy uses PPTP client with pptp-network-manager but the config info doesn't seem to be in the usual /etc/ppp/peers/ folder. The Network-manager piece probably puts it somewhere else. But there isn't an option for "persist" on the GUI like the pptpconfig used to have. I put in persist in the generic options box but had no idea if it works. Need some tests on it.

samba installation failure on Hardy

2008-09-17T06:52:00.000-07:00

The message is "subprocess post-installation script returned error exit status 102". It was done initially in Synaptic so I ran "apt-get -f install" to see the details. It looks like the culprit is

update-rc.d: warning: /etc/rc2.d/K09samba is not a link to ../init.d/samba or /etc/init.d/samba
update-rc.d: warning: /etc/rc3.d/K09samba is not a link to ../init.d/samba or /etc/init.d/samba
invoke-rc.d: dangling symlink: /etc/rc2.d/K09samba

The rc2.d and rc3.d links actually point to /samba for some reason. So I removed the symlinks and the installation went through.

The package installed was (3.0.28a-1ubuntu4.5). There's a similar issue reported on Debian sid for samba 3.0.26a-1.

The saga of upgrading Breezy to Hardy (on VMWare)

2008-07-25T12:21:00.000-07:00

It all started with me downloading UltimateLAMP, a pre-configured VMWare appliance image, to try out phpOpenChat. I did get the phpOpenChat to work (mostly) with some php script tinkering. Considering my limited knowledge on php, I'm relatively proud of myself. Even the virtual appliance hasn't been updated for almost 3 years, it still works pretty well. So I kinda thought, "why don't I make a good thing better?". My idea is to upgrade it to current software (it's based on Ubuntu Breezy) so I may even able to use it for testing some other things. Simple enough, right?

I didn't quite envision a walk in the park, but it did turn out to be close to an odyssey.

There's no direct path from Breezy to Hardy (the current Ubuntu, version 8.04 LTS). So I had to go to Dapper first (6.06 LTS). I replaced all breezy in apt source list with dapper. A "dist-upgrade" by apt-get got me to Dapper pretty easily (or I thought). But upon restart X no longer ran. Granted this is somewhat a server thing so I don't have to have X, but I just like to have a GUI to make some things easier. I went ahead to install update-manager and used it to upgrade to Hardy, thinking "I'll fix X once I'm there". It took some time to download and set up things. After a restart I found myself in an even worse shape. Hardy couldn't boot into normal mode - it would hang - so I had to drop into recovery. Then I had no network, and of course no X. To get the network to work, I realized I need to recompile vmware-tools, which contains the driver for the interface. But I couldn't: there's no gcc or make -the virtual appliance didn't include it. But without network I couldn't even download build-essential and headers. Damn! I don't want to have to make a CD-ROM...

So I grabbed the original Breezy vmdk and restarted the whole thing. This time I remembered to do snapshots along the way so it's easier to go back. I couldn't install build-essentials under Breezy, because the repository is no longer there. So I did it after getting to Dapper. For some odd reason, at presumeably Dapper level, it showed a Breezy kernel (guess either kernel image isn't there or GRUB wasn't done right). So I couldn't do "install linux-headers-`uname -r`"; instead I had to find the Dapper kernel version from another computer and used it to install the headers. With build-essential and headers I should be able to compile stuff once I'm fully upgraded. When I reached Hardy, things seemed to look better. I even had X running fine, minus mouse, which forced me to learn how to navigate Nautilus with keyboard. I thought instead of recompiling the old vmware-tools, I would just upgrade to the current vmware-tools (Workstation 6). Well, it's also because old vmware-tools reported a wrong kernel version - what the heck is @@VMWARE-release@@ - so it refused to recompile. Mounting the tools iso turned out to be a mini journey too. For some reason it didn't want to work if I make CD-ROM IDE in VM settings (missing ata driver?).

Now I got tools upgraded. Deep breath, restart. Bam! No gdm again. Still no network. WTF? It turned out that the gdm problem is just a missing symlink (/usr/bin/X11/X). But where is my network card? I updated modules.conf to reflect the correct driver, vmxnet (it was using vmnic). I could see vmxnet module loaded. But ifup kept saying "No such device". I remembered the annoying /etc/iftab Ubuntu has. With upgraded VM version, it may well have a new MAC. So I checked iftab and found out it's been replaced by an even more arcane file, /etc/udev/rules.d/70-persistent-net/rules. Opening up that file and sure enough, the old MAC was there with my eth0. And there's also an eth1 with the new MAC. The two lines are in different format. After some thinking I decided to comment out old eth0 and changed the other one's name to from eth1 to eth0. Still, there's no network after restarting networking service. So I restarted the machine (good'ol Windows habit) and guess what? All is merry again!

Hope my experience will be some help to anyone who wants to embark on such feat.

Case study: the wrong way to get new customer - Windows Live ID signup

2008-07-14T08:03:00.000-07:00

I recently tried to sign up a Windows Live ID with a non-Microsoft email. The signup page from live.com is here:
https://signup.live.com/newuser.aspx?mkt=en-us&ts=4488890&sh=7rnR&ru=http%3a%2f%2fwww.live.com%2fPassport.aspx%3frequrl%3dhttp%253a%252f%252fwww.live.com%253a80%252f&rx=http%3a%2f%2fget.live.com&rollrs=03&lic=1#ExtHelp
but the page doesn't allow you to use email other than @hotmail.com and @live.com. However when you place the cursor in the box and click the "Get help with this" context link on the right, you'd notice the instructions do say "–Or– Select Use your own e-mail address." Well, something isn't in sync here. Then, if you follow the "Windows Live ID sign up page" in the instructions, it actually brings you to an otherwise identical signup page, which does have the "Use your own email address" checkbox.

Per the spirit of a good conspiracy theory enthusiast, we'll have to rule out the chance of this being a mere overlook. :-) Apparently it's an concerted effort on steering to Microsoft services with perfect defendability on not acting like a monopoly. Well, to an honest consumer, it still feels like a trick. Either play like a whole-heartedly open-standard vendor, or be forthcoming about not being one. Either way I'd like the big M$ a little better.

umpnpmgr.dll

2008-05-29T06:32:00.001-07:00

A Windows XP SP3 laptop got stuck during boot with message "umpnpmgr.dll is a bad image" from services.exe. The file seems to be the Plug-n-Play driver. When I renamed the file using safe mode, it would say "couldn't find umpnpmgr.dll" and refused to boot into even safe mode anymore. So I got a hold of an old copy of umpnpmgr.dll from a WinXP SP2 computer and put that on using safe mode (after I renamed the bad copy back). This solved problem.

There's one thing I'm not sure: the booting issue didn't occur right after the installation of SP3. The only thing I can remember between then and now is that I put on a Motorola DCT62xx driver , so I could record stuff off it.

BitTornado error

2008-04-29T10:29:00.000-07:00

It couldn't start a torrent. The error was "path xxxxxx disallowed due to security reasons." The path seemed to be a file in the torrent package, starting with ._.

The error was solved by editing /usr/lib/python2.4/site-packages/BitTornado/BT1/btformats.py:
reg = compile(r'^[^/\\.~][^/\\]*$')
changed to
reg = compile(r'^[^/\\~][^/\\]*$')
The dot was removed.

The regular expression seems to be used to disallow any file name starting with /, \, ., or ~. That's understandable. By removing the dot it would allow files starting with dot. I'll just add it back after I'm done with this particular torrent.

Woes with Catalyst driver

2008-04-08T14:09:00.001-07:00

Since I upgraded to a 24" LCD, WMP has had some crashing issues when opening and closing. I never figured out what precise conditions triggered those but instead decided that it's time to upgrade the video driver - running a 24" is a heavier job (I had a 17") and might require a newer driver for my old Radeon 9600 Pro. The other issue is sometimes the monitor thought the output is 1600x1200 even the Catalyst Control Center said it's doing 1920x1200, so I got a desktop area bigger than the screen displays.

It turned out to be a mistake. The newest 8.3 Catalyst (internal 8.471) for x64 never worked right. After upgrading the graphics were significantly slower - windows would leave "trails" if you drag them. Dream Aquarium screensaver refused to run citing no graphics card support. DxDiag showed no Direct3D support. Actually if I disable hardware acceleration the graphics would speed up - what a joke! I reinstalled and reinstalled to no avail (also a pain to cleanly reinstall - had to do many manual cleanups to remove some parts: see here). Finally I did a full cleanup, including dropping into Safe Mode to remove some stubborn parts under Program Files (x86)\ATI\ACE. Then I reinstalled the old 6.3 Catalyst (internal 8.231), which is no longer available on ATI's site so I had to download it from Guru3d.com. Everything went back to normal after that.

Many said Catalyst is a little more flaky than Forceware. If what I had is anything close to common, than it doesn't take much to agree with that assessment...

UPDATE: 08/15/2008
Of course, the saga continues at some point. :-) The crash with closing some video files happened again a few days ago. But this time I couldn't boot into Windows XP anymore - all I got is a blank screen after Windows logo disappears. If I drop into Safe Mode, I could see an "ati2mag going into infinite loop" message. Since I have no usable computer anyways, I decided to do some experimenting. I removed 6.3 under Safe Mode then installed 8.3 under normal mode. This time the driver actually works! My 3D screensaver ran and everything seemed fine, except that the Catalyst Control Center doesn't run. But today I got the blank screen again after resuming from hibernation, and restarting it's the same. Under safe mode I could see besides the "infinite loop" message I also got "invalid CRT display". I remember seeing a cure for the "infinite loop" somewhere that involves using the "standard PCI-to-PCI bridge" driver for the CPU to AGP link. So I found it under the Device Manager (it's VIA CPU-to-AGP bridge) and replaced the VIA driver with the Windows one. That did it. I got my regular desktop back! Fingers crossed...

Virtual machine remake

2008-03-14T10:17:00.000-07:00

This old v4 VM has really got old since I had it imported into a VMWare Workstation 6. It's running out of space (6 GB on its WinXP boot partition) and VMW has been bugging me for upgrading its virtual hardware. The problem is that if I tried to upgrade, it gave me this "cannot be upgraded because it has a legacy snapshot". I probably forgot to flush the snapshot before I imported. The searching around for an after-the-fact fix didn't turn up anything useful. So I decided to just hack it myself.

So how did it know it has an old snapshot? Maybe it's in the .vmx file. After viewing it in an editor I found nothing. Maybe it's just those other files lying around in the same directory that prompted the upgrade converter. So I removed everything in the directory except the .vmx, .vmdk, and .nvram file. After restarting VMWare application, it happily took the order to upgrade!

With upgrading done, I set out to solve the space issue. With only a pair of 60 GB disks attached, this laptop running 2 OSes, dozens of big software, and a bunch of media file, is very space challenged. I still have some space on the other disk so I created a split (had to since it's a FAT32 partition) vmdk size 12 GB. I used a Acronis TrueImage Home 10 CD to boot the VM and clone the old vmdk to the new one. It took only 10 minutes or so. After adjusting the disk setting in the VM, it now runs with 6 GB free space on C:. Yeepee!

Last thing to upgrade was the VMWare tools. Once that's done, this VM is truly reborn!

Stupid VMWare Workstation error

2008-03-05T20:46:00.000-08:00

Every time I installed a VMWare Workstation 6 update, it rendered the program unusable under a non-admin account with a "Application failure: 0x80004003" error. It's all because the installer always sets some registry keys access rights incompletely. See here for an explanation and a batch file to fix those:
http://communities.vmware.com/message/362908#365731

Kick start a GBE

2008-02-01T12:29:00.000-08:00

The GBE card started to malfunction in my WinXP x64 machine. It constantly showed "network cable unplugged". I checked to make sure the cable, switch, and the networking stack are working fine - just by moving the connection to a spare embedded Fast Ethernet port, and everything worked. When I looked at the card Link/Act light and it didn't light up. Disabling/enabling and rebooting didn't help either.

So I opened up the box and swap the card with the Firewire card (I only have to PCI slots). Boot up and everything back to normal. The Firewire card works fine in the old GBE slot too. Well, maybe it just needs a little "kick start"...

Bloatware

2008-01-31T06:40:00.000-08:00

I was removing Adobe Acrobat Reader 6 from an old Win2000 machine. The Add/Remove Program applet showed it used 43.5MB space. The computer also has Adobe Acrobat (full version) 4 installed by the side, which only takes a meager 8.77MB. The new Adobe Reader 8 I think takes more than 90 MB. You'd wonder how much functionality stuffed in the bloated package actually gets used once in a hundred years. On-demand functionality, anyone?

The pesty ffdshow filter

2008-01-29T08:28:00.001-08:00

Now both WMP and RealPlayer crashed on me (DEP message) when playing some media files. I tried Media Player Classic, same thing. Well, that looks like a codec issue since the crash happens with all players. I GSpot the files and found out they are all DivX/Xvid coded. I found the utility InstalledCodec and ran it to id my codecs. I first disabled my DivX5/6 and Xvid DirectShow filter/drivers. Didn't help. Then I spotted the ffdshow filter for MPEG-4. Disabling that fixed the issue. The ffdshow filter is 1.0.0.1 (size 1761280) at C:\WINDOWS\SysWow64\ffdshow.ax. It might the relic of a few video converters I tried.

RM converter

2008-01-29T08:10:00.001-08:00

RealMedia V10 files are hard to deal with - tried at least 5 converters to transform them into other formats but none worked. Eventually I did find one that worked: Ultra RM Converter 3.3.0916 (older versions didn't work either).