Why is svn 14061 a good idea?

Denis Vlasenko vda.linux at googlemail.com
Sat Sep 9 17:09:10 UTC 2006


On Wednesday 06 September 2006 06:07, Rich Felker wrote:
> > > > > Ok, so this checkin replaces this:
> > > > > > -               if (fnmatch(list->data, filename, FNM_LEADING_DIR) == 
> > 0) {
> > > > > > -                       return (list);
> > > > > (Question #1: does uClibc support it?)
> > > > 
> > > > Yes, it does at least since 0.9.28 (and maybe before as well):
> > > > include/fnmatch.h at 58:
> > > > #if !defined _POSIX_C_SOURCE || _POSIX_C_SOURCE < 2 || defined _GNU_SOURCE
> > > > # define FNM_FILE_NAME   FNM_PATHNAME   /* Preferred GNU name.  */
> > > > # define FNM_LEADING_DIR (1 << 3)   /* Ignore `/...' after a match.  */
> > > > # define FNM_CASEFOLD    (1 << 4)   /* Compare without regard to case.  */
> > > > # define FNM_EXTMATCH    (1 << 5)   /* Use ksh-like extended matching. */
> > > > #endif
> > > > 
> > > > In a real world, we're using glibc or uClibc. Both support those extensions.
> > > 
> > > I'm not because glibc is way too bloated and uclibc is also somewhat
> > > bloated and lacks any hope of stable abi. :(
> > 
> > I typed FNM_LEADING_DIR into Google and the first hit was a six year old bug 
> > report showing that it's been around for quite some time.
> > 
> > I don't care about your personal library.
> 
> I've flamed plenty times before about why the non-Linux-specific parts
> of BB should build and work on any POSIX compliant system, so at this
> point all I'll do is ask you nicely, would you please support non-GNU
> systems?
> 
> > > There are other things that already need 
> > > to be under this anyway.
> > 
> > They never have been in the history of the project.  This is a strange 
> > definition of "need".
> 
> There are some GNU-isms that were disabled at my request with possible
> (albeit tiny) increases in size or decreases in performance. That's
> what I was referring to by need. I'm assuming that you would like to
> make the code more optimal for GNU systems without making it more
> unconditionally GNU-dependent than it already is. When I proposed the
> aforementioned changes in the past and they were accepted, it was with
> the assumption that, somewhere down the line, someone would make an
> option to choose between the GNU-specific code and the portable code.

I am thoroughly confused now. What's the consensus about
reverting/non-reverting?

[please send mails to me directly too folks, not thru list...]
--
vda



More information about the busybox mailing list