Mr. Moritz's 'make build' problem (Part I)

[Woodchuck]

Please don't quote the whole digest.  trim please.
Please use a real subject, so that later googles will find something.
In responding to this mail, use the subject I used so that things will
fall together properly.

On Thu, 17 Aug 2006, edgar mortiz wrote:

> I've pulled out src -stable via
> 
>    i have a folder /stuff/bsd
> 
>    cd /stuff/bsd
>    export CVSROOT=anoncvs at anoncvs.example.org:/cvs
>    cvs -d$CVSROOT checkout -rOPENBSD_3_9 -P src

I didn't do this, I extracted a virgin src from CD 3 of the 3.9
set.  I don't believe this will make a difference for this exercise,
there have not been changes since the CD that will affect the mechanics
of make build, as far as I know.

So whereas you are building "STABLE", I am building "RELEASE", i.e.
the "STABLE" on the day the CDs were burned.

> 
>     and when i build the kernel i did a
> 
>    config GENERIC
>    make clean && make depend && make && make install
> 
>    this one wen't well, as for building the binaries
> 
>    rm -rf /usr/obj/*
>    cd /stuff/bsd/src

Before "make obj" you must have done
	export BSDSRCDIR=/stuff/bsd/src
otherwise the make obj will fail.

>    make obj

makes objs under /usr/obj   About 1200 files are created.
(find /usr/obj | wc  gives 1180).

>    cd /stuff/bsd/src/etc && env DESTDIR=/ make distrib-dirs

This seems unnecessary unless you are making a "release".

I chose to mount a new (2GB) partition under /newb  and to
export DESTDIR=/newb at this point.

export DESTDIR=/newb
cd etc   # i.e. /stuff/bsd/src/etc
make distrib-dirs

This creates 687 files under /newb

now for the biggie:

>    cd /stuff/bsd/src
>    make build

it is now 0154 20 Aug...
I am running this on a dual PIII 1GHz machine with nice scsi system;
no other load.  I am running OpenBSD 3.9(stable) GENERIC.MP kernel.
OS was brought up to CVS spex a couple weeks ago.

>     then after a couple of minutes it breaks =/

I hope you realize that this is a less than optimum description of
the failure.  Next time, use script(1) or take a mouse cut&paste of
the xterm after it fails.

Well, I need to go now and use this xterm for something else.
The make build is cooking along (it's now 0220, it's in libcurses
or thereabouts I think.  I'll add on to this once it fails or
completes.

As usual this looks like it will work.  I've done stuff like this
in the past to debug newbie problems, and always it has turned out
that the person with the problem isn't telling me everything, or
has done something that he isn't reporting (like that "export
BSDSRCDIR" I discovered you kept to yourself ;-), or hasn't really done
something that he says he did.  This is where script(1) is so useful.
Also dumping the environment and a uname -a right before starting
the make build is quite helpful.  Dmesg is not so useful at this point.

read man script    Note where the man page coyly remarks that every
keystroke made will appear in the script.  Note that this might compromise
certain things you type while in script(1).  Edit the script file after
exiting script.

Dave