How to ARM Linux
August 4th, 2009
During last few years I’ve never found a good excessive topic covering most of the areas to get Linux runnnig from scratch. I remember myself doing a lot of investigation about correct building of a [crosscompiler][crosscompiler], studying [bootloader][bootloader] internals, and so on… I’m definitely sure – if I had a person that could point me in right direction, then all the development should finish much earlier.
Consider this article as my personal note and feel free using it your own way.
Posted in 
Tags: 
First libdiesel release
July 30th, 2009
I’ve just made first version of libdiesel library publicly available. I will not release the sources due to personal constraints, but you’re welcome to use compiled version. So far, there is only MinGW32 static version, but I may compile this library for other platform as well. Just leave me a comment.
Bloody Hinet SPAM
July 24th, 2009
Hinet SPAM reports
If one day God will give me his powers for at least few minutes, I’ll burn all the Hinet servers at once!
Simple TLS library
July 10th, 2009
Last few days I’ve been working on a simple TLS wrapper. So far, client side is finished and working lovely. Next task is to include server part, and when it will be finished I think I’ll make the library public.
AntiSPAM postfix setup
July 10th, 2009
Recently I’ve had discussion with one of my colleagues about postfix/sendmail comparison and it’s antispam mechanisms. I will not tell much about sendmail as I’m not so competent in it, few years ago all my servers migrated to postfix, after my ex-coworker have send me his postfix setup claiming it’s real-spam-proof.
Postfix configuration is more understandable (well, for me at least), and I can fine-tune it better than sendmail. In result, my colleague was really surprised when I had show him my postfix logs with 100% spam filtration (for at least last 2 weeks there was no single spam mail that where able to breach through) and asked me to mail him my setup or post the configuration somewhere.
Una Concert 2009.05.23
May 29th, 2009
Latvian National Theater Riga, Latvia, 2009.05.23
A message for those technical geeks, who may track this blog:
Sorry, I didn’t found appropriate place yet where it is possible to upload videos without size and time limitation. Consider this as offtopic 
libpqtypes
May 21st, 2009
In recent post I’ve posted part of my library, that deals with PostgreSQL data in binary format. Today, googling for some hints for implementing binary timestamps in double format I’ve found a reliable library, libpqtypes, which is doing exactly the same as mine (actually more), except it’s purpose is PostgreSQL only.
Don’t know if it’s terrible news or not, as I could adapt my code while using this library and win a lot of time. This is one of the reasons why extensive googling is required before each project.
PostgreSQL binary timestamp functions
May 21st, 2009
While writing libpq wrapper I’ve run into problems with binary timestamps, which required a lot of debugging and headaches. To avoid your own migraine, feel free to use my code below. Note, however, this code is not perfect, as it’s uses long long for timestamps. This will work if your machine is 32bit and your server was compiled with --enable-integer-datetimes (highly recommended). If not, You’re welcome to patch this code and post it here, so other people will thank you.
{
long long timestamp;
// get long long with correct byte ordering
if(!get(row_number, column_number, ×tamp))
return false;
memset(value, 0, sizeof(struct tm));
long long dDate, date0, time;
date0 = date2j(2000, 1, 1);
time = timestamp;
TMODULO(time, dDate, USECS_PER_DAY);
if(time < (long long)0) {
time += USECS_PER_DAY;
dDate -= 1;
}
// add offset to go from J2000 back to standard Julian date
dDate += date0;
// Julian day routine does not work for negative Julian days
if(dDate < 0 || dDate > (long long)INT_MAX)
return false;
j2date ((int)dDate, &value->tm_year, &value->tm_mon, &value->tm_mday);
dt2time( time, &value->tm_hour, &value->tm_min, &value->tm_sec);
// years in tm starts from 1900
value->tm_year -= 1900;
// months in tm are zero based
value->tm_mon--;
// mark this as *no* time zone available
value->tm_isdst = -1;
return true;
}
bool Database::put(struct tm* value)
{
// Julian day routines are not correct for negative Julian days
if(!IS_VALID_JULIAN(value->tm_year + 1900, value->tm_mon + 1, value->tm_mday))
return false;
long long dDate, time, timestamp;
dDate = date2j(value->tm_year + 1900, value->tm_mon + 1, value->tm_mday) -
date2j(2000, 1, 1);
time = time2t(value->tm_hour, value->tm_min, value->tm_sec);
timestamp = dDate * USECS_PER_DAY + time;
// check for major overflow
if((timestamp - time) / USECS_PER_DAY != dDate)
return false;
// check for just-barely overflow (okay except time-of-day wraps)
if((timestamp < 0 && dDate >= 0) ||
(timestamp >= 0 && dDate < 0))
return false;
long long *ts = ×tamp;
unsigned long l1 = htonl(*((unsigned long *) ts + 0));
unsigned long l2 = htonl(*((unsigned long *) ts + 1));
unsigned long long ll = l1;
ll = (ll << 32) + l2;
return put(OID_DATETIME, (char*) &ll, sizeof(long long));
}
All the functions from this code are available in postgres sources, except put and get which I will not publish due to specifics of my code. If you’re using this, you should already have your own functions for argument saving into some buffer.
Lua constants
May 5th, 2009
By Lua’s nature, it is not possible to create constants. There is a workaround, though, using metatables:
return setmetatable({}, {
__index = tbl,
__newindex = function(t, key, value)
error("attempting to change constant " .. tostring(key) .. " to " .. tostring(value), 2)
end
})
end
constants = {
x = 1,
y = 2,
}
constants = protect(constants)
print(constants.x) --> 3
constants.x = 3 --> error: attempting to change constant x to 3
Lua error message format
May 1st, 2009
I work on the library, which will use Lua. Thinking about the possibility to give a good error structure instead of plain message, I’ve found that lua_pcall has a nice feature to specify custom error handler function:
luaL_loadfile(L, "file.lua");
lua_pcall(L, 0, 0, -2);
Using this there is a way to get a chunk name and it’s error line number:
lua_Debug d;
lua_getstack(L, 1, &d);
lua_getinfo(L, "Sln", &d);
string err = lua_tostring(L, -1);
lua_pop(L, 1);
stringstream msg;
msg << d.short_src << ":" << d.currentline;
if(d.name != 0) {
msg << "(" << d.namewhat << " " << d.name << ")";
}
msg << " > " << err;
lua_pushstring(L, msg.str().c_str());
return 1;
}
Work’s like a charm, except one thing. This is the output from this error handler:
Why? Oh, why the Lua guys are putting assembled error message with chunk name and line number if I use my own error handler? IMHO, this is wrong. There is no other way to give a user nice message except parsing. Actually parsing of such message is easy, but.. the Lua guys have changed error message format a few times already. I’m pissed!