[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: GHC armel build

2007/11/16, Rene Wagner <rw@nelianur.org>:
> Initially I worked in qemu, but Wookey sorted me out with build machine access
> earlier this week.
Nice of him. I've had to build most of it on simulators and my cellphone.

> (#if defined __arm__)
> byte swapping occurs in a private copy of gmp which doesn't appear to be
> used.
Well spotted!
Fortunately, for the Debian build, -lgmp3 takes priority, but that
will cause breakage on other ARM EABI builds without libgmp3, and
means inserting defined(__ARM_EABI__) clauses into those tests.
Will you file a bug/patch to the GHC group?

> Martin, Wookey says you are also working on a GHC build for armel. Are
> you aiming for an unregisterised build or a full port? How far along
> have you got? Have I missed anything up to where I am now?
It's been relatively lumpy ride, as you will probably find, but I
haven't been writing notes as I go.
Yes, you do need a stage 2 compiler to be able to build the full final
Debian ghc so you are on the right track.
I don't know what hardware you have been given access to, but you will
find that the final build peaks at 280MB of VM, and unless you have
256MB+128 swap or lots of RAM the Debian build will die after churning
for three days (at 266MHz). Fortunately the armel buildds have 512MB.

For the moment I'm doing an un-r one, since so far that hasn't
required modification of the source and it will unblock the buildds
sooner, although it generates code that runs at half speed (including
of course the compiler itself). Not having a native code generator
halves the speed of the compiler again.
Whether I do a registerised one or not depends what ARM decide to ask
me to do. Until I hear otherwise I am assuming they want to follow the
shortest route to lenny certification.


Reply to: