futexes, PI and uclibc

[Rob Landley]

On Friday 31 August 2007 2:08:31 am Carmelo AMOROSO wrote:
> > At this point, it might be easier to just overwrite the current svn tree
> > with the contents of the NPTL branch.  The reason they havent' done it
> > yet is they're not quite sure what regressions that might introduce
> > (somebody has to go through the svn history since the NPTL branch and
> > merge all the important changes into the NPTL branch).
>
> I periodically try to include some changes on trunk within nptl locally
> (especially in case of bugs fixes),
> but, you're right, it seems there isn't currently so much activities on
> uClibc... is is due to summer time? holidays?

Well, so far I'm happy with the 0.9.29 release for everything except sparc.  
(I haven't finished configuring m68k or blackfin yet, and I'm just now trying 
arm big endian).  I don't use threading much and am not currently building 
anything against uClibc that uses it, so I don't have a test case for 
threading problems.

Also, tailor gave up the ghost trying to keep my uClibc mercurial repository 
up to date from svn, and the script I wrote by hand converts fine but I 
haven't finished the updating part (the "what state is each repository in and 
how do they correspond" bits turn out to be nontrivial to keep in sync if the 
process is interrupted, because telling if svn 131072 corresponds to hg 2236 
is non-obvious for something like the uClibc repository which has holes in it 
due to busybox/uclibc/buildroot/tinylogin/uClibc++ all being in the same 
repository, and don't get me started on branches.)  I haven't had time to 
finish that lately, and until I do i'm not going to test a more recent 
version than the last release (although I'm looking forward to 0.9.29.1).

I agree with Linus, who considers SVN's slogan "cvs done right" to be a 
contradiction in terms:
http://video.google.com/videoplay?docid=-2199332044603874737

(Yes, I'm trying to index google's tech talks at 
http://kernel.org/doc/video.html )

> > It seems like sjhill is maintaining the next release.  (Or at least the
> > primary developer thereof, at the moment.)  Could we bug him for a status
> > report, do you think?
>
> I think Steve is always in listening on the list...

Hey Steve: could you give us a status report on the NPTL merge?  We're 
curious.

> >> Our patches have been already sent to Steve for a review/merge work on
> >> June. I'm the responsible for the NPTL port on ST and I'd be happy to
> >> provided some helps for
> >> the merge stuff if required, or discuss about some implementation
> >> choices we did.
> >
> > There hasn't been a whole lot of discussion on this list about any uClibc
> > development recently, and I haven't noticed it on #uclibc on freenode
> > either. It all seems to be happening silently out of tree.  Probably I
> > should look at the other branch...
> >
> > http://uclibc.org/cgi-bin/viewcvs.cgi/branches/uClibc-nptl/
>
> neither there are being activities on the branch.
> Further, I recently wrote an email to Mike to get news about the plan
> for the merge... still not received a reply.

Mike isn't handling the merge.

If Mike really isn't interested in doing stable releases (or hasn't got time), 
I suppose I could cut a 0.9.29.1.  (I just checked and I still have root 
access on uclibc.org...)  I'd want to clear it with Mike first, though.  And 
I have to get my hg repository up and running so I can see what patches need 
backporting.  (And I'd have to go look at the bug list.)

And while I'm at it, I never did finish the 0.9.28->0.9.29 changelog, did I?  
Sigh...

> Carmelo
>
> > Rob

Rob
-- 
"One of my most productive days was throwing away 1000 lines of code."
  - Ken Thompson.