Re: Build failures on arm
On Wed, Jun 05, 2013 at 01:58:31AM +0400, Alexander Danilov wrote:
> On 05.06.2013 01:13, Colin Watson wrote:
> >That's the R_ARM_V4BX problem I mentioned here:
> >
> > http://lists.debian.org/debian-haskell/2013/05/msg00054.html
> >
> >It isn't a TH failure as such, just one that comes up a lot whenever TH
> >or GHCi is involved. I don't know why it happens to show up in that
> >particular place, but it's certainly possible that the missing handling
> >of this relocation would account for more than just the GHCi-related
> >failures on armel (not armhf).
> >
> >The patch in my above post should fix this, if anyone can test it. For
> >us, it's safe to entirely ignore the relocation like this because Debian
> >armel is (AIUI) ARMv4t at minimum anyway so we have no need for this
> >specialised relocation that's used for fixing things up to work on plain
> >old ARMv4. Even for upstream, I think the chances of anyone caring
> >about running modern Haskell code on a StrongARM are pretty limited
> >given that current distributions don't tend to support them.
>
> I run debian on thecus n2100, it's XScale-80219 arnv5te (armel), with
> 512M of memory, and this bug is annoying. I don't know exactly, but
> this bug may touch all armel machines.
I think you have misunderstood me.
The unhandled relocation certainly affects all armel machines. My
comment about plain old ARMv4 / StrongARM was regarding the way in which
one might handle that relocation; my patch "handles" it by doing nothing
upon encountering it, which is fine for anything actually supported by
the Debian armel port but would not be sufficient for old ARMv4
machines. But, as I said, I think that's acceptable.
--
Colin Watson [cjwatson@debian.org]
Reply to: