Fernando de Oliveira wrote:
Well I was worried to not let the 7.7 be released with a possible error,
so, decided to study it better.

On 27-02-2015 13:39, Bruce Dubbs wrote:
Sorry, *I was wrong* about telling as invalid the use of

mk_add_options MOZ_MAKE_FLAGS="-j$(getconf _NPROCESSORS_ONLN)"

in mozconfig. More disscussed below.

I.   CHOWN II.  Number of threads III. Make install analysis IV.
Conclusions

I.   CHOWN

After building as normal user:

su root

cd /tmp/mozilla-release

make -f client.mk install INSTALL_SDK=

Found *59 non-root* files.

# find /usr/lib/firefox-36.0/ -not -user 0 | wc -l 59

This is important, for security. You didn't mention it, perhaps
missed? Or I wrote too many posts today and you didnt have time to
follow them? That is the reason I wrote this lengthy one, so that nobady
would need to read the other ones.

chown -R 0:0 /usr/lib/firefox-6.0.

IV.  Conclusions

Conclusion 1: *chown is necessary*


Sorry, *I was wrong* about telling as invalid the use of

mk_add_options MOZ_MAKE_FLAGS="-j$(getconf _NPROCESSORS_ONLN)"

in mozconfig. It does get the correct value. Probably, I forgot to
save the mozconfig and it picked up the previously save value 4.

However, I just found that *if it is not defined* (in mozconfig or
anywhere else), it will work like

mk_add_options MOZ_MAKE_FLAGS="-j$(getconf _NPROCESSORS_ONLN)"

Conclusion 2: it is a good idea to define MOZ_MAKE_FLAGS somewhere,
or it will default to the available threads
(=MOZ_MAKE_FLAGS="-j$(getconf _NPROCESSORS_ONLN)".

Conclusion 3: if MOZ_MAKE_FLAGS is not used in the make install
line, it will use default or previously defined value. If you want
a install with -j1, MOZ_MAKE_FLAGS=-j1 must be exported or defined
in mozconfig or in the make install line.


Fernando, thank you for the detailed analysis.  Other than timing
issues for the book, do we really care how many threads are used
during the build and install?  Perhaps a note in the mozconfig is all
that is needed.

You and Ken did seem to care, when started the thread and you replied that

It doesn't make much sense to using the MOZ_MAKE_FLAGS variable for
the install phase.  I'd say go ahead and remove it when you update
the book.

I said that before you explained what is happening.

I demonstrated that if you build using it greater than 1, make instll
will not do what you two seemed to want.

Also, I am responsible for finding the old problem, which almost made me
give a j8 SBU as if it were j1. So I introduced them in the make lines,
or many people be using something different from what they wanted.

Also, ate that time it was wrong to keep it in the mozconfig as was our
tradition.

Tradition that we still have, it is in moxconfig for Seamonkey and
Thunderbird.

For me personally, it matters a lot! I mentioned earlier having had a
hardware problem. It was overheat. So, I decreased to -j5, after
reconnecting the HDs, hard rebooting, no ethernet working, rebooting
into systemrescuecd, cleaning the partitions and finally rebooting to
the present status quo.

I know you were having HW problems.  I appreciate whatever you can do.

Personally, I've been building a lot and haven't had time to consider email in detail. I'd like to get the tags done and then go back and consider other issues in more detail. In the meantime, I rely on you and the others to look at the more involved issues. Really though, the best way to keep me from forgetting is to have an open ticket. I always review those and will need to close them before a release.

Also, I'm building on a new machine. There are a lot of issues that I'm coming across. I was able to test bluez with a usb BT dongle I had, but now I need to go and get a dvd drive for it to test those functions. I also ran into an issue that I wasn't aware of when I installed iptables. Now I need (want) to get a router appliance to isolate my network from my ISPs router. I have some control over that, but not enough to suit me.

  -- Bruce

--
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