Re: 23.0.60 bootstrap fails for maemo

[Dan Nicolaescu]

Thomas Baumann <address@hidden> writes:

  > Dan Nicolaescu <address@hidden> writes:
  > 
  > > Thomas Baumann <address@hidden> writes:
  > >
  > >   > > Dan Nicolaescu <address@hidden> wrote:
  > >   > >
  > >   > >> Thomas Baumann <address@hidden> writes:
  > >   > >> 
  > >   > >>   > I'm trying to compile emacs23 for os2008 using the maemo-4.0 
sdk but
  > >   > >>   > somehow the bootstrap fails. I'm working with a fresh copy of 
the cvs
  > >   > >>   > files checked out today. 
  > >   > >>   > 
  > >   > >>   > The problem looks very much like the emacs23/cygwin problem 
discussed
  > >   > >>   > earlier, but I was not able to test all of the possible 
solutions
  > >   > >>   > mentioned there.
  > >   > >>   > 
  > >   > >>   > Any help is highly appreciated.
  > >   > >> 
  > >   > >> Not sure if this helps, but first try to bootstrap on the host so 
that
  > >   > >> you have all the .elc files, and then configure for the target 
again and
  > >   > >> just run make. (If the problem you are facing is the byte compiler
  > >   > >> encoding bug that's been discussed here, you might be able to get 
over
  > >   > >> it this way)...
  > >   > >> 
  > >   > >
  > >   > > I think I tried this before - configure with CHINOOK_X86 and compile
  > >   > > (works ok). configure with CHINOOK_ARMEL... and a new error...
  > >   > >
  > >   > > if i delete and recompile pre-crt0.o for the target I get the same 
error
  > >   > > as before
  > >   > >
  > >   > 
  > >   > After days of testing and trying to debug the bootstrap I finally
  > >   > compiled the byte-compile parts of bootstrap directly on the N810: and
  > >   > it works :-) Strange and not a permanent solution, though. 
  > >   > 
  > > [snip]
  > >   > 3. compilation is possible on the target itself: the build directory 
has
  > >   >    to be mounted with nfs and make has to be installed on the
  > >   >    device. Compile emacs until the first error occurs, continue on the
  > >   >    target and finally make install and build the package on the host.
  > >
  > > So it seems that building works on real hardware, but it fails when
  > > using the cross-compiler + simulator. That would point out to a problem
  > > in that setup, maybe a bug in the simulator.
  > 
  > I think the latter, as the setup was installed from scratch. Debugging,
  > however, is beyond my current debbuging skills. All I can offer is to
  > test possible solutions.

Then, probably the best option is to forward this to the simulator's
maintainers.