From owner-freebsd-hackers Wed Oct 16 04:48:35 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id EAA14470 for hackers-outgoing; Wed, 16 Oct 1996 04:48:35 -0700 (PDT) Received: from ra.dkuug.dk (ra.dkuug.dk [193.88.44.193]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id EAA14458; Wed, 16 Oct 1996 04:48:30 -0700 (PDT) Received: (from sos@localhost) by ra.dkuug.dk (8.6.12/8.6.12) id NAA21933; Wed, 16 Oct 1996 13:45:44 +0200 Message-Id: <199610161145.NAA21933@ra.dkuug.dk> Subject: Re: Linux compat issue(s) To: msmith@atrad.adelaide.edu.au (Michael Smith) Date: Wed, 16 Oct 1996 13:45:44 +0200 (MET DST) Cc: sos@FreeBSD.org, msmith@atrad.adelaide.edu.au, jdp@polstra.com, hackers@FreeBSD.org In-Reply-To: <199610161123.UAA03542@genesis.atrad.adelaide.edu.au> from "Michael Smith" at Oct 16, 96 08:53:30 pm From: sos@FreeBSD.org Reply-to: sos@FreeBSD.org X-Mailer: ELM [version 2.4 PL24] Content-Type: text Sender: owner-hackers@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk In reply to Michael Smith who wrote: > > sos@FreeBSD.org stands accused of saying: > > > > > Will your changes just reject an unidentifiable ELF binary out-of-hand, > > > or could you add a sysctl variable to determine which format to > > > presume? The latter method would be best, IMHO. > > > > Hmm, following the ELF style I guess default should always be native, or > > one could argue for SVR4 compliant (which you dont have yet). > > I'm not fond of it being variable, that will generate too much support > > steam I'm afraid. > > Native would be pointless, as native binaries will always be > identifiable. The reason I ask is that for the situation where the NO, wrong, we can't even see if the bin is native !!!! > Linux emulator had been loaded, but not the (hypothetical) SVR4 > emulator, you could default to Linux and not have to brand your > binaries, and vice versa. Hypothetical ?? I have a SVR4 emu :), I just can't/wont't release the code... > The only situation where branding would be required would be when both > emulations were active, and IMHO _that_ is likely to generate less > support steam just by its rarity. > > One could also bias the selection based on the path of the executable, > as that is visible in exec_elf_imgact. Anything under /compat/linux > is likely to be a linux binary, &c &c. That would further limit the > requirement for branding to "both emulators loaded and binary outside > of compatability tree". I would wote for ALWAYS branding the ELF files, that way there is NO doubt what sex they are, thus giving least trouble. > The last ditch "I have no idea, winging it" case could just default to > the last interpreter in the list - that would let the user alter the > default if there were more than one available by unloading and > reloading the LKM. I'd rather have it give up and say unknown binary format or something like that, or try running it as a FreeBSD ELF bin. Now, for a practical question, what should I call the little util ?? markelf, brandelf or just plan elf (fixelf sounds a bit harsh :) ) -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Soren Schmidt (sos@FreeBSD.org) FreeBSD Core Team So much code to hack -- so little time.