[EMAIL PROTECTED] writes:
> In my experience parallel builds normally work very
> well and can yield a huge speedup by having other
> compiles ready to run when one gets I/O bound, but I've
> seen it fail often enough that I'm always willing
> to back off and do a single threaded build as a
> sanity check when my parallel builds die for otherwise
> unexplainable reasons.
The most common problem that I see when doing a parallel build is that
the process of building a library with nm craps out:
P processes are compiling .o's, and due to the default make rules, as
soon as they're done producing .o's make runs "nm" to make a library.
Now you've got P processes mucking around with the same library,
writing to it all at the same time.
For me, this often yields a corrupted library.
(I have some ad hoc scripts to do something like a
gmake -j3 -k `echo *.c *.cc | perl -pe 's/\.cc?/.o/g'` ; gmake
if I really need a parallel build, but in general I don't use them for
important builds)
--kevin
--
Kevin D. Clark ([EMAIL PROTECTED]) |
Cetacean Networks, Inc. | Give me a decent UNIX
Portsmouth, N.H. (USA) | and I can move the world
[EMAIL PROTECTED] (PGP Key Available) |
**********************************************************
To unsubscribe from this list, send mail to
[EMAIL PROTECTED] with the following text in the
*body* (*not* the subject line) of the letter:
unsubscribe gnhlug
**********************************************************
