Freememapplet fixed

The free memory applet in the JWM tray was not displaying the free space properly when it got into the Gigabytes. I have fixed that, also ensured that it is not too wide for the display.

The current experimental system is that all the kernel modules are in the initramfs and these get moved to the main f.s. just before the switch_root, if there is enough space.

So, all the modules are present after bootup, but they will get deleted as free space gets tight. They will all get deleted except for those that got loaded, as those must remain for the next boot. The 'init' script will restore all the modules in the main f.s. if sufficient free space becomes available again. Well, that's the rough idea anyway.

The space taken up by the full set of modules is about 18MB (or maybe less if the live-CD is built with a smaller selection of modules) and I wanted the free memory applet to treat the space occupied by the modules as potentially free. So the value displayed in the tray is 18MB higher than the amount of actual free space. It is the pup_eventd that monitors free space and deletes the modules if necessary.

Posted on 24 May 2008, 22:11


Comments:

Posted on 24 May 2008, 23:02 by Ted Dog
LZMA again
The LZMA patches to initrd & kernel can be applied to existing kernel compile since it only affects the decompression code. We estimate quite a savings (8-10M reduced iso size) if used as you are heading. The decompression is actually faster in my testing than gzip. Compression on the other hand is a good time for coffee break.


Posted on 25 May 2008, 9:59 by BarryK
Re: LZMA again
So you could take the 4.1alpha, along with devx_401.sfs and kernel-src-2.6.25.4.sfs and apply the patches, then build a new vmlinuz, and it would work with all current modules.


Posted on 25 May 2008, 11:12 by Ted Dog
link
http://www.geocities.com/mctiew/ffw/lzma.assorted.patch9.tgz

I would think, the original compile environment would be best, I could never get kernels to match, some magic you do between raw and finish, I suppose.

Should only need to do the make bzImage & what ever steps you take (that magic) to create the initrd environment. Using lzma instead of gzip



Posted on 25 May 2008, 13:23 by dogone
Compression benchmark
Barry and Ted Dog, this will interest you. It's a benchmark comparing gzip, bzip2 and lzma speed and efficiency and includes kernel-related figures. Best take a look.

http://tukaani.org/lzma/benchmarks