jhalfs experimental and ICA
dbn.lists at gmail.com
Thu Mar 9 10:49:53 PST 2006
On 3/9/06, M.Canales.es <manuel at linuxfromscratch.org> wrote:
> Now, I have a dude, in part due that I don't know yet how work ICA/farce. When
> rebuilding chapter6 for personal purposes, the packages are installed op-top
> of the previous ones and not using the binaries on /tools at all. All files
> are overwritten by the new compiled ones.
> For ICA/parse, where/how each iteration is installed to be sure that the final
> product isn't a mess of files from each rebuild?
This is how I do it, and gsbuild does something similar if I remember correctly.
The binaries are installed one on top of the other in the / system.
However, at the end of each iteration, a copy of the relevant
directories is put in a temporary location for later analysis. This
requires a lot of space since if you did 2 build iterations, you'd
have essentially 3 copies of the system:
The actual final system in /
The iter1 system in /some_temp_dir/iter1
The iter2 (same as final) system in /some_temp_dir/iter2
One other thing is that /tools/bin (or whatever you call /tools) is
removed from the PATH for any ITER>1. It should be unnecessary at
this point and would identify any serious flaws in your system.
I should have just done this the first time, but I'm attaching the
actual functions file I source in my scripts. There is a bit more
housekeeping in the top level build script for changing the location
of the LOGDIR, STAMPDIR and anything else you don't want to clobber,
but I think you'll get the idea. Hopefully you don't get too bogged
down by unreferenced function names and parameters.
do_(ica|farce)_prep is called at the end of the iteration. This first
calls do_ica_files which copies the files to the temp location. Then
all the stripping, etc. is done. After all the iterations are done, a
for loop counting ITER up to ITERATIONS calls do_(ica|farce)_work on
ITER and ITER+1 to do the actual diffing on the temporary locations.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 5971 bytes
Desc: not available
More information about the alfs-discuss