build/installworld break under -j n for n>1

Stefan Lambrev stefan.lambrev at
Sun Jan 27 13:46:20 PST 2008


Aryeh M. Friedman wrote:
> Hash: SHA1
> Mark Kirkwood wrote:
>> Chuck Swiger wrote:
>>> While I find your confidence in doing an installworld without
>>> checking for whether the build completed sucessfully and without
>>> doing a human review to be admirable :-), perhaps consider doing
>>> this instead:
>>>    make -j n buildworld && make installworld
This will do almost the same as make buildworld installworld, because if 
make failed on the build part it will never continue with the install.
Also I do not have problems with make -j6 installworld. If I remember 
correctly the real gain is when running -j NumerOfCPUs+1, but only 
during compilation e.g. build part.
>>> ...?
>> Just out of interest, is it actually correct to so this? I usually do:
>> make buildworld
>> make kernel
>> reboot (single user etc)
>> make installworld
This is exactly what documentation suggest :)
I personally have the habit to skip the reboot to single user step ;)
But do not forget the mergemaster.
>> I'm guessing you would only do the make buildworld && make
>> installworld if you had *not* changed the kernel sources?
> It is normally safe if you update frequently (<once a day)
Changes in kernel are normally backward compatible, which means that in 
most cases
you should be able to run "old" world with new kernel.
The opposite is not true and you do not know when you will get bitten.
In all cases you should look at /usr/src/UPDATING, for special events :)

More information about the freebsd-current mailing list