Re: Time for Hurd hacking...

[Roland McGrath]

[Jeff is out of a job and has a week or three to devote to Hurd hacking.]

Well, condolences and/or congratulations, as appropriate. :-)

Unfortunately, Marcus is out of town this week.  Ordinarily I would quickly
punt to him for what needs working on soonest.  Of course, it depends
primarily on what you think are you interested in/good at.

pthreads is not something I would suggest, for a few reasons.  Whathisname
does indeed seem to be going at it, for one.  I have not had the time (or
really the inclination) to look into the code that's being worked on, so I
don't have anything specific to say.  But I am skeptical about these
"external" attempts to implement pthreads without rewriting major chunks of
libc.  I also think that, whatever the approach, it is likely to be
something that cannot be cleanly finished within a couple of weeks.  If you
have a short period to do some intensive hacking, it might be nice to
choose a small project you can take from start to finish (or close) in that
amount of time.

For any hacker with a clue, a couple of weeks ought to be more than enough
time to make rpctrace sing and dance.  We have now numerous novice hackers
who claim to want to work on it, but nothing has come out of it.  I think
you have more than enough Hurd experience to decide what kinds of output
formats and options are useful, as well as to implement them.  It's a
project I'd find dull, though.

Speaking of dull, if you would like to just do some intensive work on the
various web site contents, faqs, and so forth, that would be a fine thing too.
But I am presuming you are looking for something more hackety.

The things I have focussed on lately are not necessarily #1 priority, just
little things that entertain me in the short periods of hacking time I
have.  Take a look at TODO and tasks for what jumps out at you.  You have
more experience than I do lately as a Debian/Hurd user to know what the
priorities are for practical usefulness.

Of the various pending 500 lb gorillas, you might want to attack the
problems with the current libc.  See if you can reproduce Marcus's problem
when building from the current libc cvs; if so, debug it (gdb on ld.so); if
not, figure out how the debian glibc is coming out broken.  

There is also libio.  I think libc is all ready to go as far as the update
to libio (and a new soname), with the notable caveat that the transition
plan from libc.so.0.2 is left undetermined.  So that could be tested.
Start by building the current libc from cvs as it is and make sure that works.
Then do a build with --enable-libio, and test its "make check" and the like.
This build will probably be wholly incompatible with any other binaries,
but before worrying about that just test it and try a fresh hurd build
against it.