parallel builds revisited

[Peter Jeremy]

On 2007-Apr-13 17:43:54 +0200, "David Ne?as (Yeti)" <yeti at physics.muni.cz> wrote:
>implements it.  How many ports exist that can fail with N+1
>jobs yet cannot break with N jobs (for N > 1)?

As far as I can see, for a port to be safely built with N>1 jobs then
it needs to have proper dependency relationships in its build infra-
structure (Makefile or equivalent).  Once that has been done then the
only limit on the number of potential parallel jobs is the host system
resources (# cores, RAM, I/O etc).  There are definitely cases where
increasing N will slow down the build (due to resource contention) but
it should still complete correctly.

I can see that a port that has incomplete or incorrect dependency
relationships may appear to build correctly, most of the time, for
some values of N and fail for others.  I don't believe it is safe to
use parallel building with such a port because system load external to
the port build is likely to affect the relative timings of different
parts of the port build and may therefore affect the outcome of the
port build.  I am not sure how many (if any) ports fall into this
category (or how to identify them without manually checking their
dependency relationships).

Can anyone come up with either examples of ports that fall into the
second category above or counter-examples to my first paragraph?

-- 
Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20070413/134be992/attachment.pgp