[BusyBox] lash -c does not work

Manuel Novoa III mjn3 at codepoet.org
Wed Aug 11 17:53:15 UTC 2004


Hello,

On Wed, Aug 11, 2004 at 10:00:35AM -0700, Mike Castle wrote:
> On Wed, Aug 11, 2004 at 02:50:03AM -0500, Rob Landley wrote:
> > Have you tried uclibc?  It's by the same people who do busybox, so they make 
> > darn sure they work with each other.
> 
> I'm fully aware of the issues surrounding dietlibc and ulibc.

Actually, you probably aren't.  There was a lot of communication between
Felix and I in personal email.  A while back, I had tried to get the
dietlibc community to start doing testing and code review, and I pointed
out about 7 or 8 stdio files that contained bugs _besides_ the several
specific bugs I did report.  In the end, Felix's position was that I owed
him specific bug reports on all bugs I knew of... for a library I don't
even use and for bugs I found by just browsing the code for an hour or so.
My position was that if he wanted things handed to him on a platter, he
could contract me.  I wasn't interested in fixing specific dietlibc
bugs.  What I _was_ interested in was getting the dietlibc users to
start finding and fixing them themselves.  There really hasn't been any
contact between us since, although I think I've answered a question or
two on the dietlibc mailing list.

> > As for dietlibc, they have a history of ignoring patches from the busybox 
> > guys:
> > 
> > http://www.busybox.net/lists/busybox/2004-January/010598.html
> 
> I was already aware of that thread.  I'm actually using the printf.c patch
> from there (and, personally, I feel that the approach taken in busybox on
> the setting errors in printf.c is completely wrong anyway).  That thread is

Do you understand the purpose?  Unlike the other stdio functions, printf
can fail for reasons which do not result in the stream's error indicator
being set.  Since checks of the printf return codes are rarely done, it
seemed elegant to add a set of wrapper functions around the printf calls
which did that checking and set the stream's error indicator if a
problem was detected.  In that way, all stdio output errors could be
easily detected using ferror() and a lot of *printf return value checking
was avoided.  I'm open to alternate suggestions which give the same
error detection benefits without code bloat.

> also what inspired me to figure out that dietlibc has fdprintf() which is
> equivalent to dprintf().
> 
> I do agree with the statement in that particular message about icmp.  Has
> anyone submitted a patch to dietlibc yet?

No idea.  Surely people actually using dietlibc would be the ones to ask.

> Meanwhile, simply ignoring the problem isn't going to make dietlibc and
> busybox work any better.  But maybe trying to slowly whittle away at the
> issues, someone can make them both better.  First, I'm going to get enough
> working to make my toy systems work (and if that means using msh instead of
> ash, so be it).  Then after that, I'll try to work on the two systems,
> submitting patches to both as needed.

In the thread mentioned above, I gave a good bit of positive feedback
and stated that I'd be happy to help anyone interested in addressing
dietlibc issues.  But in the end, it is the dietlibc users that will
need to do the bulk of the work.

Manuel



More information about the busybox mailing list