RooJSolutions en RSS: RooJSolutions - /index.php 150 150 D sockets, and rooscript, roojs updates 2008-08-10 23:16:00 <a href="">Article originally from rooJSolutions blog</a><br/> <div> <p>Since a few friends complained about adding me to their RSS feed, then not actually posting anything I thought I'd post a little something about some of the recent hacks I've been up to.</p><p>See the extended version for details on</p><ul><li>Non-blocking socketstreams in D</li><li>Unix Sockets in D</li><li>Rooscript updates - System, and GDC cached building.</li><li>RooJS updates - examples in the manual</li></ul><p> </p><p> </p><p> </p></div> Walter showing off Digitalmars D cool features 2007-02-25 10:07:01 <a href="">Article originally from rooJSolutions blog</a><br/> Normally I dont just post a link, but this one is on the 'have to see' list. Walter Bright, the author of Digitalmars D, gave a talk to a C++ group about some of the cool features of D. - reserve an hour to listen to this one.<br /><br /><a href="">Video of Walter Bright at the C++ group.</a><br /><a href="">Slides of Walter Brights talk at the C++ group</a><br /><a href="">Event details</a> for reference<br /><br />Some of the stuff like scope(exit) - makes Exceptions actually usable. Some of the other stuff like templates I'm still trying to get my head round, let alone decide how the fit into writing maintainable code....<br /><br /><br /> Backtracing segfaults in a daemon in digitalmars D 2007-01-31 17:41:05 <a href="">Article originally from rooJSolutions blog</a><br/> One the projects I'm working on is a SyncML server, written from scratch in D, It's currently in testing mode, and we found that the server was mysteriously crashing. Unfortunatly, since it's threaded, and forked as a daemon, we didn't really want to run it under GDB, (and since GDB segfaults on startup anyway). we where at a bit of a quagmire about how to find the bug.<br /><br />So after a bit of searching through I came across the idea of catching the SIGSEGV signal and calling <a href="">backtrace</a> and backtrace_symbols <br /><br />This little trick can product output that looks something like<br /><pre>/path/to/application [0xAAAAAA] &lt;&lt; address of code<br />/path/to/application [0xAAAAAA] &lt;&lt; address of code<br />/path/to/application [0xAAAAAA] &lt;&lt; address of code<br />/path/to/application [0xAAAAAA] &lt;&lt; address of code</pre>which initially seemed a bit cryptic, but by putting it together with <br /><a href="">addr2line</a> can result in some great debugging information.<span style="font-family: monospace;"><br /><br /></span>This is my little backtrace logger for the deamon logger.<br /><pre>static void print_trace()<br />{<br /> <br /> <br /> void *btarray[10];<br /> size_t size;<br /> char **strings;<br /> size_t i;<br /> pid_t pid = getpid();<br /><br /> //writefln(&quot;SEG&quot;);<br /> size = backtrace(cast(void**)btarray, 10);<br /> strings = backtrace_symbols(cast(void**)btarray, size);<br /><br /> std.process.system(&quot;/bin/echo '----BACKTRACE------' &quot; ~ <br /> &quot;&gt; /var/log/myproject/backtrace.log&quot;);<br /> <br /> for(i = 0; i &lt; size; i++) {<br /> <br /> char[] line = std.string.toString(strings[i]);<br /> char[][] bits = std.string.split(line, &quot;[&quot;);<br /> char[] left = std.string.strip(bits[0]);<br /> if (!left.length) {<br /> continue;<br /> }<br /> // skip lines with ( in them...<br /> if (std.string.find(left,&quot;(&quot;) &gt; -1) {<br /> continue;<br /> }<br /> <br /> char[] addr = bits[1][2..length-1];<br /> <br /> <br /> std.process.system(&quot;/bin/echo '----&quot; ~ addr <br /> ~ &quot;------' &gt;&gt; /var/log/myproject/backtrace.log&quot;);<br /> std.process.system(&quot;/usr/bin/addr2line -f -e &quot; ~ <br /> left ~ &quot; &quot; ~ addr ~ &quot; &gt;&gt; /var/log/myproject/myproject.log&quot;);<br /> <br /> <br /> }<br /> free(strings);<br />}<br /><br /></pre>of course you need to use a few C externs to make this work:<br /><pre>extern (C) {<br /> int backtrace(void **__array, int __size);<br /> char** backtrace_symbols(void **__array, int __size);<br /> pid_t getpid();<br /> sighandler_t signal(int signum, sighandler_t handler);<br /> void sigsegv(int sig)<br /> {<br /> // reset the handler.<br /> signal(SIGSEGV, cast(sighandler_t) 0);<br /> print_trace();<br /> // really die<br /> exit(SIGSEGV);<br /> }<br /> <br />}<br /><br /></pre>and to add it to you application, stick this in main() somewhere<br /><pre>signal(SIGSEGV, &amp;sigsegv);<br /></pre>testing it is quite simple, just do this in D<br /><br /><pre>void testSegfault()<br />{<br /> <br /> class SegTest {<br /> void test() {}<br /> }<br /> SegTest a;<br /> a.test();<br />}<br /></pre>Now looking at the debug file, you can work out where it failed...<br /><pre>----BACKTRACE------<br />----805e971------ <span style="font-style: italic;">(THIS IS MY OUTPUT CODE)</span><br />_D9myproject7cmdLine6daemon11print_traceFZv<br />init.c/src/myproject/cmdLine/daemon.d:306<br />----805e46b------ <span style="font-style: italic;"><br /></span>sigsegv<br />init.c/src/myproject/cmdLine/daemon.d:121<br />----804db18------ <span style="font-style: italic;">(AND NOW FOR THE LOCATION OF THE SEGFAULT)</span><span style="font-style: italic;"></span><br />_D9myprojectfort7manager7manager7runOptsFZAa<br />init.c/src/myproject/manager.d:50<br />----805617a------<br />_D9myproject10webRequest10webRequest5parseFAaAaKAaZAa<br />init.c/src/myproject/webRequest.d:89<br />----8050c3e------<br />_D9pmyproject14myprojectThread14myprojectThread18dealWithWebRequestFAaAaZv<br />init.c/src/myproject/myprojectThread.d:331<br />----80503d0------<br />_D9myproject14myprojectThread14myprojectThread3runFZi<br />init.c/src/myproject/myprojectThread.d:111<br />----8076260------<br />_D3std6thread6Thread11threadstartUPvZPv<br />??:0<br />----a7fd10bd------<br />??<br /><br /></pre>I'm sure with some more work, you could get it to log to syslog...<br /> Autocompletion in leds for Digitalmars D 2006-12-12 10:16:00 <a href="">Article originally from rooJSolutions blog</a><br/> PHP Grammer added to leds, and how to build leds 2006-11-13 14:46:00 <a href="">Article originally from rooJSolutions blog</a><br/> I've spend quite a bit of time working on leds, specifically focusing on the autocompletion and help implementation for D and PHP, while the D is still ahead, in terms of cross file lookup, autocompletion and display of docbook comments, PHP is beginning to catch up. <br /><br />The major part of autocompletion for PHP is dealing with the grammer parser (as I blogged before that I had done a pretty much perfect tokenizer). For this I took alot of inspiration from the D parser in dante. Which used for me, anyway, a rather creative method of parsing the language into Components. <br /><br />In principle it breaks the language down into simple parts<br /><ul><li>files<br /></li><li>classes</li><li>methods</li><li>codeblock (code inbetween { and })</li><li>statements something ending in &quot;;&quot; ,or &quot;while ( ) { }&quot; (which is a statement with a codeblock)</li></ul>the parser method, basically relies on each class to find all it's subcomponents, then let's the subcomponent determine where it starts and ends. - relying on the file to store the pointer to the current position in the list of tokens.<br /><br />The result is a very simple to write grammer parser, that can be grown organically, rather than the classic grammer parser that depends on the need to fully document every pattern in the language.<br /><br />When I originally started this post, it was intended as install instructions for leds. which are included in the extended body. And also partly as a reference for me to get it up and going quickly on other machines. The documentation uses Makefiles (as you would use with C), however, after having chatted to Antonio Moneiro, the original author of leds, he had been working on a compile tool (compd).<br /><br />This compile tool negated the need for makefiles, as it basically worked out what commands to run to build an application, by just being given the files to build, binary library paths and source library paths (bit like .h files in C)<br /><br />This solved alot of the issues which the rather hacked together makefiles creates, but didnt solve the major problem. That you need to install the libraries in specific places, or modify the build command to help it find the libraries. eg. leds, uses libraries dantefw, dool, dui (GTK) and has to know where you have built those libraries, and where the source is.<br /><br />So with a little bit of hacking, I added code to the build tool, such that when it has built the library, it writes a list of file paths and dependant libraries to either ~/.compd/{libname}.compd or /etc/compd/{libname}.compd.<br /><br />Then for example when building leds, you just tell it that you need -ldantefw, it will look in /etc/compd/libdantefw, and find where you compiled the library to, and all the include paths. - hence you no longer need to specify paths anymore, and applications and libraries just build. without editing make files or using autoconf.<br /><br />In addition we are still discussing how we can make it run more like &quot;make&quot;, where it picks up a file in the current working directory, and uses that as it's arguments. so making leds would be as simple as typing &quot;compd&quot;, or &quot;compd -f leds.compd&quot;<br /> <br /> <br /> Parsing PHP in D. 2006-06-01 08:55:54 <a href="">Article originally from rooJSolutions blog</a><br/> I keep meaning to write a full log of the whole mythtv experience, as the number of hardware headaches that that has gone through is enough for at least 4 or 5 posts (including the need to rebuild from scratch this whole server a few weeks ago due to hard disk failures from overheating..). But this week, for what seems like the first time this year, I actually scraped 2 days to work on yet another new pet project. Phpmole's replacement....<br /><br />Phpmole is my lifeblood for development, when I wrote it, I added all the features that where missing from other editors, and the resulting editor made a huge difference to my productivity.<br /><ul><li>code folding<br />Although annoying at times, is a good way to dive into new code.</li><li>autocompletion<br />look up PHP functions / variable names etc.</li><li>inline help hints<br />show the signature of a php method inline.</li><li>List of open files on a left bar.<br />Since I often have 30+ files open, this is very usefull to flip between them.</li><li>Standard editor features, like syntax highlighting etc.<br /></li></ul><br />However Phpmoles code base is now pretty old, and was written before I did much PEAR work (eg. it's messy....). There are alot of design decisions that related to PHPGTK1, and the need at the time to implement file transports (ssh/midgard/file) etc. Which are now redundant as things like sshfs do that much better.<br /><br />So I've started hacking on <a href="">leds</a>, the editor I mentioned before when hacking on D. It has the benefit of being relatively small codebase wise, and very easy to understand. Let alone it's fast and runs as a binary, so I can eventually just compile the lib's and distribute them, rather than a huge array of php files.<br /><br />Anyway, the first stage in the great conquest was to teach leds a little about PHP. So starting with Zend's lexer, I hand crafted a reasonably complete lexer in D, (it ignores variables in quoted strings, and heredocs), but should be enough to grab defined classes/methods/vars etc. and work out where to put the folding... <br /><br />While at present, it's has a few in-efficiencies, as the design was a little organic. It is however quite an interesting way to build a generic language parser and only took 2 days to write and test.. <a href=""></a><br /> D compile time learning... 2006-03-23 16:32:26 <a href="">Article originally from rooJSolutions blog</a><br/> Random notes during compiling:<br /><br />method Argument missmatch: <br /><span style="font-family: courier new,courier,monospace;">mime/Document.d(264): function dinc.mime.Source.Source.skipUntilBoundary (char[],uint,bit) does not match argument types (Source,char[],uint,bit)</span><br /><br />method overloading -&gt; although you can overload a method within the same class with different signatures, in the extended class, creating a different signature will not work unless there is a matching definition in the base class.<br /><br /><br /> More DigitalMars D - finding a string in a stream 2006-03-15 10:14:53 <a href="">Article originally from rooJSolutions blog</a><br/> The three good ways to learn a language:<br /><ul><li>hack on some existing code</li><li>write a simple program from scratch</li><li>port some code from another language to the one you want to learn.</li></ul>Well, this week I though I'd have a go at the third. Picking something that was well writen to start with, I decided to use <a href="">binc</a>, an extremely well writen imap server which is writen in C++, and see how it converts to D.<br /><br />Rather than attacking the core imap bit, I decided to start with the MimeDocument decoding part. something relatively self contained, and conceptually quite simple. Most of the Porting involved bringing together Classes that had methods defined in multiple files (as seems common with C++), and merging them into nice classes in D.<br /><br />While most of it will probably end up untested until it's all ported, one single method stood out as a good simple test of working with D. - Searching for a string (or delimiter) in a stream.<br /><br />Obviously, one of the things that happens with an imap server, is that it has to scan a email message, and find out how what makes up the email (eg. attachments, different mimetypes and how they are nested. A brute force approach would be to load the whole message into memory, and just scan through looking for the sections. However, since email messages can frequently be over 5Mb, It's obviously horribly inefficent. So the existing code used a simple C++ method to search for a delimiter.<br /><br />Hit the more link for another simple tutorial...<br />