Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 17 Mar 2004 09:09:25 -0800
From:      Drew Tomlinson <drew@mykitchentable.net>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: Failure At Build (?) Stage When Making Ports?
Message-ID:  <40588645.6070108@mykitchentable.net>
In-Reply-To: <4051DC00.1000009@mykitchentable.net>
References:  <4050C378.7040802@mykitchentable.net> <200403120114.00830.danny@ricin.com> <4051DC00.1000009@mykitchentable.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On 3/12/2004 7:49 AM Drew Tomlinson wrote:

> Danny Pansters told a big fish story including the following on 
> 03/11/2004 4:14 PM:
>
>> On Thursday 11 March 2004 20:52, Drew Tomlinson wrote:
>>  
>>
>>> I'm trying to do things the "smart" way.  I have two machines running
>>> 4.9.  Instead of keeping a ports collection on both,  I have and update
>>> the collection on one named blacklamb.  Blacklamb runs Samba 2.2.8a_1.
>>> I created a samba share called "ports" and pointed it to /usr/ports.  I
>>> then used smbfs to mount "ports" on blacksheep (the other machine) at
>>> /usr/ports.  Here's the relevant portion of /etc/fstab from blacksheep:
>>>
>>> blacksheep> cat /etc/fstab
>>> # Device                Mountpoint      FStype  Options         Dump
>>> Pass#
>>> //<account>@blacklamb/ports /usr/ports smbfs rw,noauto             
>>> 0      0
>>>
>>> On both machines, I edited /etc/make.conf to set 
>>> "WRKDIRPREFIX=/var/tmp"
>>> so each would use it's own disk space when making ports.
>>>
>>> Ports build without error on blacklamb, the machine that has the ports
>>> tree locally but when building on blacksheep, they always fail.  I've
>>> read the Porter's Handbook to see what I could figure out.  I think
>>> blacksheep is failing at the actual "build" stage.  I posted a complete
>>> build log of an attempt to build the bacula client at
>>> http://drew.mykitchentable.net/Temp/blacksheep_bacula.txt.
>>>
>>> All attempts to build ports on blacksheep fail at this same point.  I
>>> have no idea what to check next so any help would be greatly 
>>> appreciated.
>>>   
>>
>>
>> Looking at the output I noticed it was in 
>> /usr/var/tmp/usr/ports/theport/workdir/and/all/that.
>>
>> Was that the intention? Or is there some symlink /usr/var to /var, 
>> considering the WRKDIRPREFIX. Since the 'depend' step seems to work 
>> fine, your method in itself can't be wrong. Maybe you need to use 
>> hard links for some reason that I can't quite grasp (yet)...  
>>
> I set my system up with a 100MB / and the rest as /usr.  So yes, tmp 
> is symlinked to /var/tmp and usr is symlinked to /usr/var.
>
>> If you're doing something with symlinks, I'd look there first for an 
>> explanation why the build fails. It says 'can't read makefile'. You 
>> should be able to find out which one that is (have it build locally 
>> without cleaning).
>>  
>>
> However, the machine that has the ports tree install locally 
> (blacklamb) is setup the same way.  A smaller root partition and the 
> rest as /usr with symlinks.  Ports build OK on blacklamb using both 
> make and portupgrade.
>
>> Amplify this if you also have a symlink /tmp to /var/tmp (consider 
>> scripts using '../..' in paths etc). Does the same thing occur 
>> without using portupgrade (which certainly uses /tmp), e.g. just make?
>>  
>>
> Problem on blacksheep occurs with both make and portupgrade.
>
> I'm going out of town for the weekend but will experiment when I 
> return.  Maybe in the meantime someone can either confirm or deny your 
> suspicions.  :)

I set WRKDIRPREFIX = /usr/var/tmp but the same error occured.  Any other 
ideas?

Thanks,

Drew



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?40588645.6070108>