Le 19/11/2014 17:53, Fernando de Oliveira a écrit :
On 19-11-2014 13:41, Pierre Labastie wrote:
Le 19/11/2014 17:18, Fernando de Oliveira a écrit :
On 19-11-2014 12:47, Pierre Labastie wrote:
Hi,
Currently, the instructions for firefox tell to pass
MOZ_MAKE_FLAGS='<number>' to the "make -f client.mk" commands.
It seems to me that it should be:
MOZ_MAKE_FLAGS='-j <number>'
And actually, when passing MOZ_MAKE_FLAGS='3', I get:
[...]
make 3 -j4 -C /sources/firefox/mozilla-release/firefox-build-dir
make[3]: Entering directory '/sources/firefox/mozilla-release'
make[3]: warning: -jN forced in submake: disabling jobserver mode.
make[3]: *** No rule to make target '3'. Stop
make[3]: leaving directory
'/sources/firefox/mozilla-release/firefox-build-dir'
/sources/firefox/mozilla-release/client.mk:392: recipe for target
'realbuild' failed
[...]
Also, according to the doc, it is possible to add a line:
|mk_add_options MOZ_MAKE_FLAGS="-j4"
to the mozconfig file (not tested). Wouldn't that be better?
It seems you missed:
https://www.mail-archive.com/[email protected]/msg01641.html
Ooops, sorry, forgot about that. So just consider the first part of the
message (replacing "<number>" by "-j <number>").
I use -j4 for the book, so it doesn't defaults to make -j8.
Sorry Fernando, but I think you are missing the point: in the book it is
written (litterally):
MOZ_MAKE_FLAGS='/|<number>|/'
it should be:
MOZ_MAKE_FLAGS='/|-j <number>|/'
MOZ_MAKE_FLAGS='<number>' (without the "-j") does not work for me.
I hated doing this modification, when there was also the optimization
problem yo be solved. It was done after many experimentations, not
because I "just wanted that". And I don't consider a solution, but a
workaround for a big in the code.
I tried 1, 2, 4, in mozconfig, always got it running make -j8.
Never thought about 3 or other odd integer.
number of cores+1... Anyway, not a big deal. In the doc, they say to go
up to 1.5*number of cores.
Think I read that and all indicating that mozconfig should be honoured.
Believe I read also discussions demonstrating that that cores+1 was
frequently a very bad choice. I read lots of things, due to this bug.
I was just explaining why I tested -j 3 (and signaling some silly
upstream suggestion)... I understand that finding such a bug triggers a
lot of documentation reading!
I've tried by hand, in script.
I have found that variable in several places of the source code and in
some places it seemed to ignore some values and force others, but, IIRC,
it was always for debian.
Thanks for all that. I am just building firefox right now for installing
icedtea-web, to try it with OJDK 8. It is a long overdue work, and I am
in a hurry to finalize it...
Maybe, I'll come back to that issue later to see if it is possible to
find a workaround.
Workarounds I already gave two: yhat one in the book or export ... and
unset ..., as I explained in the thread.
Don't understand what you mean.
I was thinking of a sed to get back the former behavior, but you are
right, that is useless, since you have found what to do! Let's hope the
former behavior will come back at next update.
Regards
Pierre
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
