[blfs-dev] #3982: Firefox-23.0.1/Xulrunner-23.0.1

Bruce Dubbs bruce.dubbs at gmail.com
Mon Aug 19 07:42:17 PDT 2013

Fernando de Oliveira wrote:
> Em 18-08-2013 22:13, Fernando de Oliveira escreveu:
>> Em 18-08-2013 17:46, Bruce Dubbs escreveu:

>>>> Command attached before was broken incomplete. I hope you all do not
>>>> mind, but I am sending another one, from a build log, deleted previous
>>>> and following lines, leaving some which seemed relevant.
>>> This all seems like it's consistent.  Perhaps a note about the RAM/swap
>>> would be appropriate.
>> OK. I will do it.
> I have a problem with this:
> [attached pdf, comparing six systems]
> I have omitted the system having 16GB of RAM.

Yes, the system will need more RAM/Swap when using a high -j value.

> Have rebuilt yesterday, to confirm numbers, with the developing system
> with increased memory. It seems total memory size was about the value
> used when linking libxul, when I read while building yesterday, so it
> failed for a small amount, 200MB was already used, before build started.
> But why the i686 systems do not fail? Without much experience with
> x86_64, builds, may I assume that binaries are larger for this architecture?

My benchmarks some time ago showed that 64-bit binaries were about 15% 
larger than 32-bit binaries.  I would think that would hold in memory too.

> Is it reasonable to insert a note for xulrunner's page mentioning it is
> only with x86_64 architecture that the problem occurs?

I'm not sure about that.  My first thought is that the different 
compilers may need significantly different amounts of memory.  I'd just 
note that "On some systems, the RAM/Swap combination needs 6 GB of 
memory available."  or something similar to that.  It's hard to pin it 
down more than that.  Extra swap today doesn't really cost the user a lot.

   -- Bruce

More information about the blfs-dev mailing list