On 08-11-2014 13:46, Fernando de Oliveira wrote: > On 08-11-2014 13:36, Bruce Dubbs wrote: >> David Brodie wrote: >>> On 08/11/14 12:07, Fernando de Oliveira wrote: >>>> >>>> Subject is seamonkey, but error is in *firefox-33.0.3*. >>>> >>>> Parts of build log attached. >>>> >>>> Left configuration and install parts. >>>> >>>> Optimization error starts at line 571, key word is "Executing" at the >>>> beginning of line (search ^Executing). >>>> >>>> Your (Bruce, who asked, and anybody else, please) help is much >>>> appreciated. >>>> >>>> I will start by comparing python versions in the two systems: >>>> >>>> LFS7.1 host (SVN-20120311) (error occurs) >>>> >>>> LFS7.6 (no error with enable-optimize >>>> >>> >>> The following relates specifically to Firefox 32.0.1 (and Seamonkey >>> 2.29), but may well still apply to 33.0.3. >>> >>> There is apparently a gcc compiler issue causing the Python script, >>> 'packager.py' to fail while it is precompiling a list of Javascript >>> scripts, using Mozilla's (internal) js shell, 'xpcshell' - essentially, >>> it (i.e. xpcshell) falls over about half way through the list, as your >>> log demonstrates. >>> >>> Now as I understand it, for gcc 4.9.x, certain 'undefined' behaviour has >>> been changed, in particular, the handling (signed) integer overflow, so >>> that any program that was tacitly relying on a certain behaviour will >>> have problems, which could very well be what is happening here (i.e. >>> some counter is overflowing). >>> >>> See: https://bugzilla.mozilla.org/show_bug.cgi?id=1047803 for some >>> discussion. >>> >>> (In other words, it is a gcc problem, not a Python problem) >>> >>> Using ac_add_options --enable-optimize="-O2" (rather than >>> --disable-optimize), appears to cure the problem for both packages. >> >> That's interesting. On my 686 system I still have gcc-4.7.0. I'll try >> to build it with the full --enable-optimize and see what happens since >> the bug report indicates the problem started with gcc-4.8. >> >> If that doesn't work, I'll try --enable-optimize="-O2" > > I have updated gcc in the ode machine on 28-Jun-2014 08:51: > > $ gcc --version > gcc (GCC) 4.8.2 > > I have updated LFS-7.6 to gcc-4.9.2, the it might have been this update > the responsible for the better behavior of firefox-33.0.3 there, because > today I diffed it to 3.0.2, and essentially only windows, not linux, > parts changed, what surprised me. > > Now I am finishing the update of the old machine to gcc-4.9.2, after I > read David post..
Success!!! Thank you very much, David and Bruce. David, once more you give me a very appreciated help. So, removing the fix for to disable optimization in the update I did yesterday was correct. However, will wait for you, Bruce, so, in the text we can offer a choice between upgrading gcc or using --enable-optimize="-O2", instead of disabling it completely. It is impressive the difference in performance (start time, downloading page time, ...). If I understand correctly, you are testing FF. I will try to build SM with optimization, now. After all done, will fix the book (SM, hopefully, and FF). For MOZ_MAKE_FLAGS (other thread, only Firefox), I will use MOZ_MAKE_FLAGS='-j<number>' make -f client.mk MOZ_MAKE_FLAGS='-j<number>' make -f client.mk install INSTALL_SDK= and explain "wher you choose <number> as the number of parallel jobs you want to use, usually less than the number of threads of your processor. If MOZ_MAKE_FLAGS is not defined, it will try to use the maximum allowed in your system.". And remove from mozconfig. Having tried this and export, this is easier for scripting, because it is unnecessary to unset it. So, no need for two more commands or thee if in the script there is an option for DESTDIR. Bruce, would you prefer doing the modifications yourself, or fix after I edit the SM and FF pages? -- []s, Fernando -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
