Date: Mon, 15 May 95 15:03:50 MDT From: terry@cs.weber.edu (Terry Lambert) To: sef@kithrup.com (Sean Eric Fagan) Cc: bde@zeta.org.au, freebsd-hackers@FreeBSD.org, henrich@crh.cl.msu.edu Subject: Re: login.c environ=envinit ? Message-ID: <9505152103.AA10970@cs.weber.edu> In-Reply-To: <199505151815.LAA11739@kithrup.com> from "Sean Eric Fagan" at May 15, 95 11:15:47 am
next in thread | previous in thread | raw e-mail | index | archive | help
> I'm all for having environment variables accessible through the kernel > (/proc/<pid>/env, anyone?), but the main reason *I* haven't done so > so far is that there *is* code out there that knows how to manipulate > the environmnent variables, and assumes things about the global variable -- > or even that there is one ;). > > If terry wants to implement it and try it out for a while, I'll be glad to > help him, but I wouldn't want it in the mainline of code until, say, all of > ports has been built to use it. > > (Wow... code [of a sort] from terry. Will wonders never cease? 8-)) I've had it running (logical names and variant symbolic links) for quite some time on my 1.1.5.1 system at home. I am still reluctant to upgrade it to 2.0 mostly because of the disk slice code (I'm still running my own idea of what that should be like and it works except in the OnTrack case, but I can ignore that). I'm a little hesitant to risk 4G of disk contents (even though it is backed up doubly to QIC tape), both because my house isn't well net-connected and because I don't have a CDROM to fix things in case of problems. Not to mention that I'll have to throw out or rewrite a lot of file system and networking and internationalization code to get back to where I am today. As far as all of the ports being built to use it, there's little difference between a (3) routine and a (3) routine that calls a (2) routine to do its thing. Everything built with a shared library will just work. I'm using my own pre 2.0 shared library system, and the exec changes are pretty massively mixed up with each other (my stuff doesn't have a crt0.o, per se) and other changes to proc for shared descriptor tables and other crap (like there isn't a vnops tiny-table in my vnodes for socket support, open files are reference counted instead of duped in the system open file table, and the offset is maintained in a per proc struct that's intermediate to the file table, etc.) . The proc and exec changes would be the logical place to start hacking, but any diffs I could give would be 100%. Let me get yet another machine at home before I'm expected to start giving out diffs. I'm reminded of Jeff Goldblum doing brain surgery at the start of "Buckaroo Banzai". 8-). What does Rod sell ASUS dual pentium boxes in tower cases for? Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9505152103.AA10970>