On Sat, Jul 30, 2016 at 08:55:40PM -0500, Douglas R. Reno wrote:
> Michael Daum wrote:
> > Hi developers,
> > 
> > I have two hints that may be useful and mentioned in the BLFS Book (systemd
> > version):
> > 
> > 1) Building Firefox-47.0.1 in parallel mode failed for me. The building
> > process stopped several times. A few times because of missing files, e.g.
> > "BUILDID.h not found" if I remember correctly. I had to build with the
> > "-j 1" flag to succeed. Then everything worked fine. I dont think that my
> > sytem is too old or slow (Core i7-3930k), sources were compiled on a SSD.
> > 
> > So the usual hint on the firefox build section that parallel builds may fail
> > on some systems for some reason might be enough here.
[...]
> 
> As for the firefox issue, I have no idea. I'd like someone else to chime in
> here. I only have 4 cores on my i5-2400, and I am a bit far from being able
> to compile it at this time.
> 
I've just taken a look at what I did on my haswell i7 (I record the
 -j value in the 'stamps' I write when a package is updated), but on
ff-47.0 [ I didn't see any reason to update to 47.0.1 ] I used -j4
on both 7.9 and the more recent svn build.  I guess that I probably
used -j4 so I could measure it for BLFS in the current style.  I did
build 47.0 several times whilst sorting things out, and I'm _fairly_
sure that some of those were probably -j8.  But I have no evidence
to back that up.

I do think that an Ivy Bridge machine (iN-3xxx) is more likely to
have problems than a recent machine, because over the past few
months, for me parallel builds with jN where N is greater than 1 have
usually failed on less-well-endowed systems (A10 Kaveri before that
died, i3 SandyBridge).  But I'm sure I've alosnoted at least one
similar build failure on my i7 haswell, I just don't remember which
package.

So, for the moment, _I_ think the 'usual hint' is sufficient.  Of
course, ff-48 will likely be released in a few days and I guess that
means there is a possibility that everything will have changed (I
don't normally build the ff beta's, so I have no idea how much will
change).

What does puzzle me is Michael's "a few times" : if a parallel make
fails more than once for me (sometimes, only once) I drop back to
 -j1.  I can see that for tedious packages (firefox just about gets
into that category, webkit and qt more so) on a well-endowed machine
I might try -j4 before -j1. But I only have the one machine which
claims to have more than four cores: on my (dead) A10 with an SSD I
used to use -j5 instead of -j4 and I think that where -j5 gave a
build failure, -j4 did the same, but on the i7 haswell which now also
has an SSD I have not been motivated to try -j9 instead of -j8 :-
diminishing returns.

ĸen, currently updating my font details, and also my build scripts
to record what other people have committed, so still a long way from
doing fresh builds.
-- 
`I shall take my mountains', said Lu-Tze. `The climate will be good
for them.'     -- Small Gods
-- 
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page

Reply via email to