Just to let you know I haven’t forgotten [well actually I had forgotten, until those accursed reminder dialogs started popping up] and will be posting new versions of JujuEdit, BookReader and eLibrary in the next couple of weeks.
Mid-June is my approximate deadline for these. If for some reason I fail to meet this I will make available previous versions with the expiry dates disabled.
Just saw something in some of Raymond Chen’s sample code that made me go "Wha…?!?" in a good way. It’s a line where he initializes a classic Windows struct, MONITORINFO, which like a lot of windows structs starts with a cbSize value [to hold the structures byte size— used as a kind of version/error check].
Normally in such instances I would write [with a sigh]:
Whereas I noticed that Raymond wrote:
which achieves exactly the same effect incredibly neatly! I can’t believe I was unaware of this technique [specifically the fact that even though only the first value is specified, the entire struct in initialized to zero]. Brilliant! And you don’t even have to know the name of the stupid cbSize member!
BTW: I know I’m not the only one who uses [until now] the tedious long form— If you are a C/C++ person I hope this tidbit can save you valuable seconds as well. And if you are a legend who already uses the short form, please don’t think less of me for not seeing it myself… ;)
Since this blog is supposed to be about my software, I thought maybe I should post some info about development status.
Comments & suggestions [+ extra bug reports] are welcome!
Daylight Savings Time.
It seems that every time we pass in or out of this period, I have some kind of hassle. On previous occasions I would simply slap myself on the forehead, recompile and upload my website and move on, making a mental note to "watch out" for time zone issues in the future.
Not this time. This time I decided to find out why this keeps happening. And to my surprise, I think it may not be entirely my fault!
Like a lot of software, my HTML editor checks file modification times to determine when pages need rebuilding. Unfortunately it seems that some incorrect documentation has led me to intepret the values returned by a certain low level system call incorrectly.
The documentation for the _stat(…) function implies that the returned st_mtime member contains a UTC time value, with the sample code using a ctime(…) call to convert it to a string representation, and ctime(…) is designed to take a UTC value as input.
From my own testing however [on an NTFS file system] it seems that st_mtime is pre-adjusted to the local timezone, meaning that passing it to ctime(…) will result in an incorrect time value, representing neither UTC or local time, because it will be performing the adjustment twice.
Returning such an adjusted value in st_mtime seems particularly inappropriate since the raw value will acually vary depending on the current time zone setting of the local machine as well as the Daylight Savings status! This seems to be the root cause of my files getting out of sync when DST changes occur.
I am now in the process of fixing my code so that I get UTC values instead of "adjusted" time when querying file properties… hopefully this will improve some of the issues I’ve been having.
Have been doing some serious code cleaning recently, fixing an age old link problem that had been preventing all of my software compiling with the multithreaded DLL versions of the C++ run-time libraries.
It was all down to a single stupid #pragma directive ticked away in a header file, which was actually excluding the required LIB file [although there was no obvious sign that this is why the lib was not linking properly — and I’d forgotten it was even possible to do that with #pragmas]. I think I added the directive years ago to avoid some MFC link error… clever me.
Now that I can finally link my software against the DLL runtime libraries, I can shave between 50 and 90K off the average executable size… but I probably won’t bother since it will be at the expense of a new DLL dependency. Still, it’s nice to get it sorted and at least have the option.
*
BookReader is in the process of getting UTF-8 support, which is a nice thing, and I’m testing a version of eLibrary which uses the aforementioned index files now available from Gutenberg.
*
JujuEdit… soon soon soon… I’ve got to get some new user configuration options in there; The damn thing doesn’t even have the standard Tools->Options property page, with everything still handled using sub menus with checkable items. Yuk.
Have made available for download a "compiled" version of my JujuScript calculator, which got a bit of a fix-up with some recent JujuScript changes.
JujuScript is drifting further away from pure C/C++ syntax, and it’s kind of a relief because it was never 100% compatible anyway. It’s nice to not have to conform. Still there are areas I have to be careful about, such as operator precedence, where it would be kind of insane to break with C standards.
That said, I do think that C/C++ operator conventions* are annoying for the following reasons:
I guess the precedence issue probably had something to do with the evolution of C, and is not the sort of thing you can change without breaking a lot of people’s code, but the ^^ operator has never been used in C, so adding it for C++ would have affected no one.
The only reason I can think of for leaving out logical XOR is to discourage people from relying on implicit casts to bool [and in fact it’s possible that the only reason we even have logical AND and OR is for their short-circuiting behaviour].
Here are some other things I think are strange about C/C++…
__________
* Java uses the same conventions.
eLibrary is a project started a few years ago, which never worked all that great, in part because of the crappiness of Project Guteberg’s database; basically I had the program downloading one giant directory listing of raw filenames, and one [inconsistent, apparently hand edited] listing file, and attempting to correlate them in order to create a usable index of titles and filenames. Given the unfriendly source data, there was only so much I could do, and so many books were missing or incorrectly indexed, and in some cases whole slabs of text files got mis-labelled as "audio books". By the time that started happening I had kind of given up on ever getting it right; because of the supreme crappiness of this approach [which took much-o tweaking with hideous regular expressions] there wasn’t all that much I could do to fix it.
Not any more!
Finally Project Gutenberg offers a machine readable index file [possibly they always had one, but if they did it sure was well hidden]. So now eLibrary can get a revision which will enable it to take full advantage of the amazing power of XML and not screw up author and volume names! [ eg in the current version I note that "The Bible, King James" was written by "Book 1, Genesis" ].
There are still other problems with PG texts though, namely shitty formatting, but I can moan about that in another post.
Ok, so i know Wordpress is not necessarily bug free, but I really wish its XMLRPC was a little more robust.
Jujublog used to be my personal weblog, but then because it was less and less about Jujusoft and more and more about me avoiding work, I decided to move it to intepid.com.
Now I am reviving the JujuBlog name with an eye to providing more software/business related info here. Whether or not that will pan out I’m not sure, as I’ve never tried keeping more than one blog before.
We shall see…