[uClibc] uClibc++ update

Garrett Kajmowicz gkajmowi at tbaytel.net
Sun Sep 5 15:32:14 UTC 2004


Version 0.1.2 is out.  Lots of good stuff.

http://garrett.dyndns.biz/uClibc++/download.html

From the changelog:

-Completed all core code for set, multiset, map and multimap.  Rewrote to use 
inheritence to cut code duplication
-Added more algorithms
-Now compiles the stlport 1.0 testsuite (some minor changes to not test 
extensions needed)
-Does not yet link stlport testsuite - missing implementations of a number of 
functions
-Added sort algorithm - using a bubble sort for now.
-sort_stable will likely remain bubble sort for some time (It is both stable 
and in-place)
-sort will likely be converted to a heap sort eventually (It is in-place, but 
non-stable)
-List sort is a bubble sort.  Might change later (insertion sort?), but it is 
difficult to work with list limitations
-Split iterator code into two sections to cut down on dependency hell
-Compiles ncurses c++ library and test application

I'm still not able to run gcc 3.4 on my system.  I do not have the space to 
install another uclibc environment (buildroot), and I need the one I already 
have.  If you have gcc-3.4, please try and compile the libraty and send me 
all of your error messages - I'll fix everything for the next release.

The reason that I am planning on using bubble-sort (or it's relatives) and 
heap sort is that they are in-place sorts.  Since we are aiming for embedded 
systems, memory should be considered to be at a premium.  Given that, 
in-place algorithms are the way to go.  The extra memory they require is 
constant (and usually just a single spare element).  Heap sort it pretty fast 
and in-place, but is unstable, and therefore isn't suitable for stable_sort.  
I'm nor sure if there is an algorithm which is better suited for stable and 
in-place conditions.

Please let me know what you think of all of the above.

-	Garrett



More information about the uClibc mailing list