~ overflow ~

Problem:

Since i’m broke, i need to get this type of jobs in order to make some dough, it’s not that i don’t like it, but it’s good to have a challenge ahead.

So i got 2 old  machines to reinstall the OS, might look a simple task, but it wasn’t. Both machines didn’t regonize the keyboard, one of them, required an adapter from the ps/2 to the big ps/2 which i don’t know the right name.

So i was stuck on the POST, “Keyboard not found press any key to continue” — hahahah — also one of the machines got a battery malfuncion, which means that once it’s unplugged it’s BIOS will reset to factory defaults.

Solution:

First i tried and old `big ps/2` keyboard i have here, i though that my keywboard might be not working right. FAIL.

Then i tried an alternate ps/2 hook i had. FAIL.

On one of the machines i was able to bypass the POST, however, the machine’s devices weren’t configured, so i couldn’t boot. Although the keyboard was working, misteriously, NOT on the POST.

So i decided to do a hard reset on the BIOS, for the first machine it instantly worked, i got the keyboard with the adaptor working just fine, so i could boot the system up and reinstall the OS. Now for the second one, i couldn’t get into BIOS Setup, keyboard wasn’t being detected at all. half FAIL.

I have this ps/2 -> usb thingy, it allows me to plug a ps/2 mouse and a ps/2 keyboard and converts it to a usb port. So i tried this and it, finally worked. In this case i had to enable legacy usb keyboard support on BIOS, in order to have it working after POST.

I also had a spare motherboard battery (from an 386 board) which worked just great on this , also crappy, old computer, so no more BIOS resets.

Hope it lasts for more than a month so i don’t have to redo it for free!

Problem:

How to display detailed information about each table on a database?

Solution:

show table status;

Source:

MySQL

Problem:

Which method is faster? MySQL? MySQL HEAP? Direct File access? Semaphore?

Solution:

First a little explanation:

This test will incrase a variable/value from 0 to 10000, I did this test to figure out what’s the best method to do a heavy usage counter.

MySQL: will do a simple update query on a table, update tbl set counter=counter+1 limit 1

MySQL HEAP: will do the same of MySQL, however using a heap table which is stored directly on computer memory, many people claims that this is the fastest method.

Direct File: will store the variable on a file that is readden then written everytime a change is needed.

Semaphore: will store a variable straight into kernel shared memory (those variables will also be avaliable to programs external to apache/php)

Regular Variable: just for comparsion propouse, this test will show how much time it would take to incrase the counter into a regular php variable, not shared or anything.

Test Setup:

Athlon +2200 256kb cache

1GB RAM

Fedora 10

MySQL 5.0.77

PHP 5.2.9

Apache 2.2.9

Results: (test type (numbers added): how much time it took – how many times can be done in a day)

Regular Variable (10000): 0.003432035446167s – 251745651684.61/day
Semaphore (10000): 0.8582170009613s – 1006738387.8812/day
File (10000): 124.96164608002s – 6914121.4692927/day
MySQL (10000): 46.476830005646s – 18589908.130461/day
MySQL Heap (10000): 24.242248058319s – 35640259.018945/day

Source script

Conclusion:

As stated on the tests, semaphore is way faster than any other method, it might be a bit tricky to work with but it really did the job well, MySQL HEAP, was the second, althought it’s about 24 times slower, might be a good solution for you if you need to work with databases, however, MySQL HEAP has some limitations that might be a issue. Regular MySQL also shown itself fast, but about 2 times slower than heap. The file method could be faster on a SCSI or SSD, since i don’t have any of those i will stick with my slow SATA disk, however, i don’t think that even on a fast SSD this would be faster than semaphore, one simple reason is that the I/Os would be so many that they would limit the speed and viability of this idea, and also consume massive system resources.

* A note about semaphore: This method uses a shared part of memory, you can raise the limit of shared memory depending on your needs (check shared memory configuration link), note that if you’re going to use this method by many different processes you need to build up a queue system in order to make it work right, php has all the functions able to do it.

* Note 2: Semaphore stored values will not go away if php or apache crashes, they are stored straight into kernel shared memory, unless the whole machine crashes/reboot it will still there.

Links:

Shared Memory Configuration

PHP Semaphore Manual

MySQL Table Storage Engine Manual

Problem:

Let’s assume you want to build up fake statistics on google analytics using a php script.

Solution:

You just need to input the google analytics UA code and you’re ready to go:

Problem:

I was looking for an API to do searches by google using a custom made php script, so i could display the results the way i want. I found out about a google search api that work with SOAP, however it’s been deprected and they only have the useless AJAX option, which don’t really serve my propouse since the searches would be client-side processed at AJAX, and i needed them server-side processed.

Solution:

This might looks like a hack, but there’s nothing much else we can do, although, for me, it was much easier than doing the research on the original api and finally finding out it was deprected. This PHP script will do a regular request to google, then it will process the data using DOM, cut it removing google ads and other crap, and finally return only the urls of the search. This will also make you able to search on country specific google, just change the domain and filter to get the results you want.

So here’s the PHP source that don’t depend on google happyness to work:

* Note: You need DOM Object support in order to run this script; This example has been developed to accept input from the CLI, you can easly change it to work inside your script or/and accept $_GET or $_POST.

Problem:

Some services keep dieing and there’s nothing much you can do about it, since they don’t restart automatically.

Those services could be httpd/apache, mysqld, sshd, etc

Solution:

This little script will do the job:

pstree | grep httpd | wc -l | awk ‘{if ($1 == 0)  system(“service httpd restart”) }’

in this example i’m checking if httpd (apache) is running.

broken down, pstree will list all your running processes,

grep httpd will search for “httpd” string on pstree,

wc -l will count the words from grep httpd, basically will return 0 or 1 if something was found,

and finally awk will check if wc -l is = 0 , meaning that the service / process in question is not running, then will perform the required action, on my example, service httpd restart which will restart apache.

You may also check it by port level as suggested by HTNet :

netstat -ln | grep “:80″ | wc -l | awk ‘{if ($1 == 0)  system(“service httpd restart”) }’

You could build up a little .sh file and put it on crontab to check it every minute or whatever you like.

Problem:

How to make windows update agent (the one on tray) to run when you want?

Solution:

net stop wuauserv (not necessarly needed)

wuauclt.exe /detectnow

be patient, it might take a while for it shows up on tray and start downloading

Sources:

Microsoft KB

Problem:

How to build vidalia from source at FC6 ?

Solution:

Ignore the INSTALL file, apparently it’s outdated and qt is no longer the same they link there, you can install all the needed packs from yum:

yum install qt4 qt4-dev* nas tor privoxy qt cmake

Then it’s just build from source:

cd /path/to/vidalia-src

cmake .;make;make install

done!

Source:

Vidalia Project Trac

Extra:

For me, tor didn’t worked at all it kept connecting forever and couldn’t build a node list, i figured out that the latest tor engine for fc6 installed by yum was 0.1.x which didn’t work with the actual network anymore, so i had install tor 0.2.x from the source, after that it almost connected, tor kept building a nodelist forever, so i checked the message log and figured out that the machine clock was wrong, in order to fix that, i had to sync it using ntp.

Problem:

How to call a function from a different frame/iframe ?

Solution:

It depends where the function is, when it’s on the main frame, or if you’re calling a funcion that’s on _top from inside a iframe:

top.function();

If you’re calling a function from a frame to another:

parent.frames['frame_name'].function();

or

parent.frames[frame_number].function();

Problem:

I had a bunch of files but i didn’t knew which ones were doubles and which ones were unique, they had different names and belong to different subfolders.

Solution:

A PHP script that reads and do a md5 hash on every file then do a comparsion.