On Sun, May 25, 2003 at 11:43:44AM +0300, Ruslan Ermilov wrote: > Bear in mind the conditionals of make(1) and sh(1)? Neither of which install files. I hope ;) > But you don't often know if installworld attemps to install it > or not, if it wasn't here: > > src/sys/boot/i386/loader/Makefile: > > beforeinstall: > ... > .if !exists(${DESTDIR}/boot/loader.rc) > ${INSTALL} -C -o ${BINOWN} -g ${BINGRP} -m 444 \ > ${.CURDIR}/../../forth/loader.rc ${DESTDIR}/boot > .endif > > > Other than the general case of using install -C, can you > > point out any specific examples? I thought that was the point > > of mergemaster; to manage elements of the distribution where > > simply overwriting would be inappropriate. > Nope, mergemaster(8) deals with updating of /etc only, unless > I'm horribly mistaken (I don't use it personally, but pretty > much do what it does manually). What you just said implies that installing world+kernel, then running mergemaster, does not update every part of a FreeBSD base system installation, modulo ports and make.conf knobs. I knew that already about /stand. I didn't realise it extended elsewhere. Perhaps because I lack install-fu. I confess that my Astonishment meter just twitched a little. Any reason why /boot/loader.rc *shouldn't* be under mergemaster control? At any rate, I'd expect to catch-and-document special cases, if only through a find-and-diff. I might just have a crack at this during the week. I have a deep sense of it being needed. Joshua -- Joshua Goodall "tea makes itself" joshua_at_roughtrade.net - Ana SusanjReceived on Sun May 25 2003 - 01:06:04 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:09 UTC