[uClibc] uClibc-gcc Did I miss it?

Manuel Novoa III mjn3 at codepoet.org
Wed Nov 19 18:31:08 UTC 2003


Hello,

On Wed, Nov 19, 2003 at 12:24:03PM +0200, Jaco Greeff wrote:
> Hi,
> 
> Ok, I must admit, I haven't been closely reading the uClibc mailing lists as I 
> should. (Too many lists, too little time.) I've just tried to update the 
> Mandrake package to the new uClibc, since it is used to build busybox which 
> in turn is used by mklivecd (http://livecd.berlios.de) in it's initrd. Which 
> worked quite well up till now.
> 
> So now only I notice that the uClibc-gcc toolchain is gone. Following the 
> changelog, I see it has moved into the debian folder (0.9.22) and 
> subsequently removed (0.9.23). Am I wrong in saying that this approach to 
> building with/using uClibc as an "inferior" library (could not think of any 
> other words) is no longer supported?  

We no longer support the gcc wrapper.  For reasons, see Erik's post.
http://uclibc.org/lists/uclibc/2003-October/007315.html

> In this specific environment the use of a buildroot is not an option, since we 
> are only building one package (that I'm 100% aware of) with uClibc. (Well, 
> that is apart from the users, you never know quite what they all get up to ;) 
> I feel quite stuck, last time I tried to get busybox to build with dietlibc I 
> didn't get quite far enough. I understand that the push is towards buildroot 

I've been waiting for Felix to send me a patch to fix one issue for
past 2 months.  He doesn't seem very interested, in spite of his
rant in the changelog of the last dietlibc release.

Also, if you are interested in using dietlibc, you might want to
take a look at my posts to their mailing list over the past couple
of months (  http://news.gmane.org/gmane.linux.lib.dietlibc/  ).
I've been trying to motivate them to start doing more testing and
internal code review... with limited success.

> and towards embedded systems, but there are valid uses for it in a mixed 
> environment.

Not at all.  It is simply that we can no longer reliably convince
the stock glibc-based toolchains to do what we need them to do,
and are therefore forced to build uClibc-specific toolchains.

> So I see one of three choices:
> 
> 1. Dump uClibc in favour of something else.

You're options are quite limited there, and you'll eventually wind up
running into the same kinds of problems.

> 2. Port uClibc-gcc from 0.9.2[12] back into 0.9.23 and hope for the best.

For your stated limited needs (I'm assuming _static_ busybox), you'd
probably be safe.

> 3. Keep on using the (outdated) 0.9.21 version in Mandrake.

If the warpper is working for you for 0.9.21, it will probably work
for you if you port it to 0.9.23.  It's just that we no longer support
it.

> None of these sound very appealing.

For your needs, probably (2) is best at this time.

Manuel




More information about the uClibc mailing list