Date: Fri, 28 Jun 1996 21:24:11 +1000 From: Bruce Evans <bde@zeta.org.au> To: jkh@time.cdrom.com, rgrimes@gndrsh.aac.dev.com Cc: current@freebsd.org, nate@mt.sri.com, scott@statsci.com Subject: Re: Building inside of /usr/src? Message-ID: <199606281124.VAA24857@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>> If you read the comments above this section of code, and have worked on
>> asymetric AMD managed user home directory systems, and used pmake
>> in parallel mode on these systems it would be clear as to why it if
>> preferintial to use $PWD if $PWD infact resolves to the sameplace as
>> getcwd().
>But PWD can't be trusted, as we've already seen. How would you
I think PWD can be trusted to be NOT set except at the top level. It
isn't supported by /bin/sh, and it shouldn't be exported, so for
`cd $subdir; make', PWD is never set. Thus PWD is very rarely set
except for developers who cd to a bottom level directory and invoke
make there. Then it is convenient for ${.CURDIR} and ${.OBJDIR} to
be short paths through symlinks.
>invocations of the build return totally different obj directories and
>the only reason this didn't become a problem before was because the
>"window" for failure was narrower - you had to have a bogus $PWD at
>the time you built the links rather than just at any time.
I think the obj links were canonical paths because pwd was used to
create them. Perhaps this is why pwd was used - ${.CURDIR} might have
been non-canonical.
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606281124.VAA24857>