Tuesday, August 15, 2006

Asrock wasted precious several hours of my time

Motherboard manufacturers compete in terms of no of functions they have. In the process of cramming all kinds of functions even in low end mobos, they just don't pay attention to the quality aspect of what they do.

After several months of running ok, a computer I had it assembled (AMD Sempron, Mobo: AsRock K8Upgrade-VM800) for a friend of mine starting giving BSOD, slowness etc. Not having faith in anybody locally in his location to fix, I had it shipped to Bangalore to have a look.

Boot up at my location, it came up properly. Looking at the event log it has a bunch of following errors several times since the date he complained:

Event Id: 26
Source: atapi

The driver has detected that device \Device\IdeIdePort0 has old or out of date firmware. Reduced performance may result

Event Id: 51
Source: disk

An error was detected on device \Device\Harddisk0\D during a paging operation

Event Id: 9
Source: atapi

The device \Device\Ide\Ideport0 did not respond within the timeout period.

XP didn't let me finish a Acronis True Image backup complete. It would fail close the end of the backup job. I bought TuffTest just for this and ran the CD-ROM based diagnostics. The surface test gave disk time out errors after 20 mins or so.

Suspecting disk first, moved the disk alone in another AMD Sempron based machine with different Asus motherboard, the disk worked properly and the Acronis Backup running on top of Bart PE just went fine as well as much faster than the time it took (40 mins) before it failed on the other machine.

Now that disk is ruled out and backup is done moved the disk back to the original machine to further troubleshoot. While playing with BIOS settings, I bumped into the SATA Operation Mode option which was set to "RAID". The XP was using it as a IDE disk though. Changed it to "Non-RAID" at BIOS, booted XP again came back fine. It still took 40 mins to do Acronis Backup but went through just fine. So is the TuffTest surface analysis test.

Works but slow. Found that this is due the transfer mode for disk is running in PIO mode only though the disk is capable of DMA mode. Tried all kinds of solutions like messing with registry changes etc to let XP recognize it as DMA disk. So, the first conclusion to make is, though operating as RAID at BIOS, it still allows the OS to recognize the disk as IDE albeit at a much slower PIO mode. Installed the AsRock supplied disk drivers. Still no luck.

I found a RAID driver in their website and literally has to reinstall over the current XP to change the disk driver to the RAID one. All of a sudden, the speed of disk became DMA mode. So, the Mobo guys only allow full speed DMA mode under RAID driver only.

By not doing either of the following, they really wasted a lot of my time.


1. RAID is the default so force the XP to recognize this as a RAID device and ask for the driver during install time. This would have avoided me unknowingly thinking that disk is IDE and everything is fine.
2. If allow the dual mode of recognizing RAID as IDE when only one disk is present, then let the drivers access it in full speed DMA mode.

Saturday, January 28, 2006

Java Docs Search made easier

About a couple of years ago, as we were developing with J2EE and whole bunch of open source libraries (perhaps some 20 in number) one of the things I observed to be clumsy is locating Class names or method signatures while coding. At times you only remember a part of a class name or a method name. You end up having to go to the web to look for it. Each library's JavaDocs are in different places and it does waste you time.

There was a need for a tool to search across multiple JavaDocs. Upon Googling at that time, we found that there was a desktop based utility available (I don't recall it's name right now) to search Java Docs across multiple libraries or JDK versions. The beauty of it is that it takes the actual Java Docs files as it is and indexes them. The negative side of it is it needed to be installed in each developers PC and we had over 30 of them in the team.

I felt that it would be a good small project for someone to make the above as a web based application to save the deployment effort each time a library changes. When we were welcoming a couple of new colleagues, this project was a obvious choice for one of them when we needed him to get familiar with Struts.

Hence the "API Search" tool was born and used internally fairly regularly. From the beginning, I wanted this to be open sourced as a way of paying back to the community and also for collective improvement and maintenance of it. For one reason or the other (though not directly due to my previous employer as I did have the approval to contribute to FOSS wherever I see it fit) it wasn't done and I also left my previous employer last year.

My friend and ex-colleague Karthik attended the foss.in with me in last November. He was very much impressed by the FOSS approach there and went back with the renewed vigor and one of the first thing to come out of this fresh energy is releasing of APISearch as a open source tool under GPL license. I am going to join this project as a developer myself and contribute anything I can to this. I am certainly proud of Karthik for this.

Go check it out. Please give feedback there or comment in this blog and we will take this further.

Return of Waterfall Methodology

Many of the thousands of readers of this blog(ok. ok. two or three readers) know that I really enjoy attending technical conferences as a way to keep myself up to date and often motivate(ok. almost coax them) others to do so.

After returning to Bangalore, I do miss all the good stuff on the conference front that goes in the US. As much as it is very expensive to attend a conference in the US, here is one conference I have decided to spend the hefty $5000 and attend. I seriously recommend you check this out.

I like what Andy's blog says about it:

For those interested in returning to Real programming, you might be interested in attending Waterfall 2006, a new conference designed to celebrate the joy of slower, sequential, plan-based development. It'’s sponsored by the United States Postal Service, the SEI, and Lager'’s Pub of Canton MD.

Be sure to click on the "Register Now" button, it really captures the flavor of this... er... event.