plan9port on AIX

[Russ Cox]

forwarded from private mail

Subject: plan9port on AIX

------------------------

From: Benjamin Huntsman <BHun...@mail2.cu-portland.edu>
Date: Tue, Jul 14, 2009 at 8:41 AM
To: r...@swtch.com


Russ-
  Since I work on AIX for my job, and I wanted to have rc and some other plan9port items, I did a wee bit of porting work... just enough to get the parts built I needed.  It should be a good starting point for full AIX support, though.

  Would you be interested in incorporating some of it (cleaned up a bit) into the distribution?  Or would you only be interested in accepting a 100% complete port?

  The biggest pain is the fact that 'uname -m' on AIX returns the machine ID (serial number) instead of the processor type.  Therefore, all the sed scripts that transform this don't work.  I got around that by hard-coding my build machine's serial number, and having the sed scripts directly replace it with something useful.  Perhaps you'll have a better solution...

  I also got a working port for Tru64 which I have at home, though I did that about a year ago and haven't kept it up to date.  I'd have to look through it before letting it out into the wild...

Thanks!

-Ben

----------
From: Russ Cox <r...@swtch.com>
Date: Tue, Jul 14, 2009 at 9:01 AM
To: Benjamin Huntsman <BHun...@mail2.cu-portland.edu>


I'm happy to incorporate pieces as they are ready.
Manageable sized chunks are best.

I am experimenting with reviewing and accepting
changes using codereview.appspot.com, a web app
for viewing and commenting on patches.

Assuming you have checked out the Plan 9 tree using
Mercurial, you can try this by downloading
http://codereview.appspot.com/static/upload.py

Then run:

cd $PLAN9
path/to/upload.py -r rsc --cc codebot --send_mail filenamelist

It will diff against the most recent checkin in your
Mercurial repository.  If you have your own commits
then you need to add --rev log-id-of-plan9port-base
to get the right patch to come out.

That uploads them to a server and sends mail about
them to both of us.  Then if I have comments I'll reply
and you can edit the files and re-run

path/to/upload.py -r rsc --cc codebot --send_mail -i issue-number filenamelist

The new -i flag takes the number that codereview.appspot
assigned the first time.  It's a bit too recipe-heavy right now
but I hope to integrate this process into hg itself via an
extension or two.
Ouch.  Is there a uname flag that does the right thing?
Maybe we can run (uname; uname -m) pipe through
the sed and pick off AIX in the first half.  Does AIX run
on more than one architecture?
Sounds great.  I doubt much has changed in the last year.

Russ

----------
From: Benjamin Huntsman <BHun...@mail2.cu-portland.edu>
Date: Tue, Jul 14, 2009 at 9:59 AM
To: Russ Cox <r...@swtch.com>


That all goes back to the "hg doesn't work over ntlm proxy" problem I mentioned in another e-mail.
It'll be a beautiful day when it does, and I'm sure it will get it given the popularity it seems to be enjoying lately.  For now, I'll tar up the changes and submit from home...
'uname -p' prints "powerpc".  OpenSolaris says that this is preferred over 'uname -m' also.  Maybe we should just switch to that across the board?  I verified that this prints out usable info from AIX, Solaris, Mac OS X, and Linux.  Here is what gets printed by each OS after running 'uname -p':

Solaris: i386
Linux:  i686
AIX: powerpc
Mac OS X/PPC: powerpc
It did at one point in history, but not anymore.  AIX's 'uname -p' should always yield "powerpc", even on POWER systems.  I verified this on several of our systems, from POWER4+, POWER5, and POWER6.  I hard-coded "powerpc" into a script or two, because that's a safe assumption on AIX.

I'll package the Tru64 stuff up too, but I'll have to poke through it first.  I'll try to send you parts of the AIX stuff tonight.

Thanks much!

-Ben

----------
From: Russ Cox <r...@swtch.com>
Date: Tue, Jul 14, 2009 at 10:09 AM
To: Benjamin Huntsman <BHun...@mail2.cu-portland.edu>


On Tue, Jul 14, 2009 at 9:59 AM, Benjamin
I thought that was only hg authentication.
This is an unauthenticated hg session
and should be fine.
Sounds like a good plan to me.
Thank you.  Do you mind if I copy this thread to plan9port-dev?

Thanks.
Russ

----------
From: Benjamin Huntsman <BHun...@mail2.cu-portland.edu>
Date: Tue, Jul 14, 2009 at 10:12 AM
To: Russ Cox <r...@swtch.com>


Unfortunately, the issue is that hg has to authenticate to our proxy, before it can even make a non-local network connection.  So, even though your hg repo is not authenticated, I still can't get past my proxy.  Oh the joys of the corporate world!  But, like I said, I'll just deal with it by carting my work home at night.  It's livable for now.
Please do.  That's where it belongs anyway.  I just thought I'd start with you to get a general idea of how complete my patches needed to be before submitting.  I'll pass bits and pieces to you as I get them working.

Thanks much!
-Ben