*BSD News Article 1647


Return to BSD News archive

Xref: sserve comp.unix.bsd:1680 comp.protocols.ppp:561
Path: sserve!manuel!munnari.oz.au!samsung!sdd.hp.com!think.com!cayman!brad
From: brad@Cayman.COM (Brad Parker)
Newsgroups: comp.unix.bsd,comp.protocols.ppp
Subject: alpha of ppp-1.1; bug in locore.s/ovbcopy()
Message-ID: <BRAD.92Jun23132656@haiti.Cayman.COM>
Date: 23 Jun 92 17:26:56 GMT
Sender: news@cayman.COM
Followup-To: comp.unix.bsd
Organization: Cayman Systems Inc., Cambridge, MA
Lines: 44
Nntp-Posting-Host: haiti


I have a new version of the PD ppp for sun's and (now!) 386BSD. It's not
a major improvement over the "ppp-1.01" except that it now works on 386BSD.
I tested it against Morningstar PPP 1.2.1 and 1.3.  Seems to work ok.
Based on the milage, I would call this an "alpha" release.

It is available for ftp from ftp.cayman.com as pub/ppp/code/ppp-1.1.tar.Z

It has some minor bug fixes, a couple of new features and a new man
page.  I changed the release dir structure around, separating the sun
stream code and the bsd interface code.  The sun stream code includes
the recent "rwall" bug fix (I have not tested this yet, however).  The
top level makefile has some notion of installing the source.

Interestingly, the major problem I had porting to 386BSD was due to a
bug in locore.s/ovbcopy(); It appears the case of 

	"ovbcopy(cp, cp-1, len);"

does not work.  It produced repeated sequences of 3 bytes in my
decompressed TCP packets ;-( Needless to say, the if_ppp.c code now
has a nice "dumpbuffer" routine for mbufs and buffers ;-) In my
travels, I would say this only affects VJ tcp header compression and
IP options.  Not being good at 386 assembler, I did not debug the
ovbcopy (but I plan to try ;-)

This release has the call to ovbcopy() in slcompress.c commented out and
replaced with some C code.  If someone has a fixed ovbcopy(), please send
me one.

Also, the new session code in 4.3+ "setsid()" is more "POSIX compliant"
than Sun's setsid() (which "does the right thing", but is not strictly
POSIX).  I believe adding a fork solves this.  The 386BSD setsid() appears
to be strict POSIX.

-brad

ps: Everyday I go to the mail box (a.k.a. agate.berkeley.edu) and hope
for a letter (a.k.a. 0.1). ;-)

--
A metaphor is like a simile.

Brad Parker	Cayman Systems, Inc., Cambridge, Ma.	brad@cayman.com