[klibc] Development forever
olecom at flower.upol.cz
Thu Sep 6 21:56:29 PDT 2007
Hallo, Peter. Cheers, developers!
I'd like to join this effort, in case i'll useful with my crazy stuff.
First of all, please consider subscribing this ML to Gmane interface.
It would be great! Especially, if you will provide full-text archives
for importing. (All that web archives suck, you know. Small and big
archives are most easily handled with news server/reader.)
Secondly. Did you started klibc as `for of dietlibc'?
Why i'm asking, not for trolling or that kind stuff. I just want to know
your opinion about inclusion some comprehencive tools in addition to
dash and kinit.
For example how klibc relate to posix utils by Jeff Garzik?
I'd like to have sed in setups like klibc. This is because i can write
quite rich applications only in dash/sed. This requires BRE support,
as opposed to just shell patterns. I'm about to start TUI system based
on it and TERM=linux. Latter may be expanded via Linux development of
vt, i guess. Because current prehistoric set of capabilities is just
dead and stupid. Careful (re)design of such thing can lead to
comprehensive framebuffer support even on console based apps in the
end (i.e. world domination).
I saw your changes to dash, like timeout support. Why is that? Bash
is good example of NIH stuff not to follow.
After quite some time of crazy shell scripting i realized, that most
fundamental thing, like select(2) just don't exist in shell. Realizing
that, together with readings of POSIX requirements for basic tools,
i've concluded, that all that stuff just... You know, give the word...
Maybe some other parts looks good etc, tools part is very astonishing.
And i saw, that you've introduced perl recently to klibc sources.
Maybe this is also because basic tools are just crap.
Anyway. I'm doing my homework. I've rewrote most of the `test` built-in
in dash, with current more than 50% win in size. In summer i tried to
contact Herbert, concerning some patches (in Debian's BTS) and ideas
to improve. He still didn't reply. While i see his activity in LKML, i
don't see such in dash.
How do you handle this? Also, do you know about sharing of code with
original NetBSD shell?
I want to get rid of `make` as phenomenon. Using non POSIX, but
adopted, `-nt` option in `test`, main frature of `make` transforms
test source.c -nt build-time-stamp -o ... && $cc
I also plan to do optimization for efficient processing of long list
Having kind of tcc, maybe very small set of tools will be capable of
building full featured programms and kernels in future.
Finally. Don't know what your plans are: good luck!
More information about the klibc