OSLib and ELF

[Tony van der Hoff]

On 13 Apr at 0:08 Stewart Brodie <stewart at metahusky.net> wrote in message
<gemini.jgeqxu00br7vb03qs.stewart at metahusky.net>

> John Tytgat <John.Tytgat at aaug.net> wrote:
> 
> > In message <gemini.jgeev7002fu6z03qs.stewart at metahusky.net>
> >           Stewart Brodie <stewart at metahusky.net> wrote:
> > 
> > > > The -j make option.  Basically doing more than one build step at the
> > > > same time.  Keeping multicore/cpu happy.  It roughly cuts build
> > > > times in 2 or 3.
> > > 
> > > Does that *really* matter all that much?  Last time I cross-compiled
> > > OSLib, it only took 45 seconds or so for a full build from clean but,
> > > admittedly that was probably 7-8 years ago, but I can't believe that
> > > it's got slower in the meantime, has it?
> > 
> > Actually on my machine (AMD 64 X2 4400+ based):
> > 
> > real    4m5.554s user    3m1.671s sys     0m53.119s
> 
> How bizarre.  I remember being shocked that it had built so quickly - I
> actually assumed that it hadn't built correctly at first, but it had.  I
> suppose it could have been because I was able to use objasm to assemble
> it, and I'm assume that you're not using it.
> 
Stewart, as I recall, you never built the entire project, instead only
building the library binary, and the (C?) headers. The thing that takes the
time is building the stronghelp manuals. I think the test suite was after
your time, and I'm sure you didn't build the distribution directories. 

That would certainly account for the difference in timings.

> Idle curiosity really: how long do native builds take nowadays?  It used
> to be 45-50 minutes on a SrongARM Risc PC, IIRC?
> 
> 
No, several hours, again for the entire project.

> > > I don't understand why you can't use the same Makefiles for
> > > cross-compiling and native compiling - is it not just a case of a
> > > different CC and LIB macros?
> > 
> > Exactly, that's where I want to go.
> > 
> > > I thought I'd submitted all those changes to the project when I did
> > > them, which would be at least 7-8 years ago now.  Perhaps it was just
> > > the patches that meant you could compile up defmod itself for Linux or
> > > Solaris. Sorry, if that's so - I no longer have any RISC OS source
> > > trees :-/
> > 
> > I can't answer that.  Your defmod changes are certainly there,
> 
> It's a shame.  Looks like that work was lost then :-(
> 
I incorporated everything you submitted, which in fact led to DefMod2. Your
makefile changes were subsumed into the current GNU makefiles. I'm not aware
of any work being lost, and I'm pretty certain you gave me everything you
had.


-- 
Tony van der Hoff        | mailto:tony at vanderhoff.org
Buckinghamshire, England