[klibc] Patch queue pushed out

Thomas Bächler thomas at archlinux.org
Tue Aug 11 10:17:47 PDT 2009

maximilian attems schrieb:
>> http://repos.archlinux.org/viewvc.cgi/klibc/trunk/klibc-Kbuild.patch?revision=12688&view=markup
>> klibc-Kbuild.patch just omits a few executables in the build procedure, 
>> and replaces a few static ones with shared ones, nothing fancy.
> remember to have asked that for hpa.
> could you split that up in two?
> so that the shared one could maybe reach upstream if hpa acks?
> the other part is arxlinux specific indeed.


>> http://repos.archlinux.org/viewvc.cgi/klibc/trunk/klibc-fix-2.6.28-includes.patch?revision=22942&view=markup
>> klibc-fix-2.6.28-includes.patch: This is an ugly hack to make klibc even 
>> build with merged x86_64 and i386, klibc was never adjusted to this 
>> unification and thus always failed.
> as hpa is x86 maintainer, let him speak a word on this,
> looks hackish.

It IS hackish. It was a quick solution so I could build klibc against 
2.6.28 when it came out. I assumed that this would be fixed upstream 
soon, so I didn't bother fixing it properly. Who knew that this still 
wouldn't be fixed when 2.6.30 came out.

>> http://repos.archlinux.org/viewvc.cgi/klibc/trunk/kinit-fix-init-cmdline.patch?revision=41787&view=markup
>> kinit-fix-cmdline.patch: Ever since we used klibc, "init=/bin/sh" 
>> stopped working, as kinit always passed the kitchensink to init 
>> (/sbin/init doesn't care, /bin/sh does and thus fails). The way I 
>> understood it, everything _before_ the first -- should be passed to 
>> init, while the rest is for kinit (combined with what was on the kernel 
>> commandline to build some mount options and more). This simple patch 
>> implements this behaviour.
> we had a similar trouble with initramfs-tools and upstart.
> some init implementation do cleanup on ENV they receive other not.
> to me it feels right to assume they don't. so ack too.
> also as archlinux is the distro actualy using kinit,
> guess your point on kinit itself has lots of weight.

Provided that many people like to use "/bin/sh" as their init 
implementation to fix boot problems, we must assume that the init 
implementation is stupid.

This patch has a weird side-effect though: Usually, sysvinit overwrites 
its own commandline to look like "init [5]". Since I've applied this 
patch, it just says "ini". This is just ugly, not a real problem.

> could you prepare a git tree somewhere with corresponding acked patches?
> hpa likes when the subject of the patch is prepended with [klibc]
> and has meaningfull desc (but guess you are fluent with git am or
> apply).


git:// and http:// links for cloning and pulling can be found at that 
link. The repository has a "for-upstream" branch with the mentioned patches.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
URL: <http://www.zytor.com/pipermail/klibc/attachments/20090811/a29d036c/attachment-0001.bin>

More information about the klibc mailing list