Re: [Chicken-users] [Chicken-announce] CHICKEN 4.10.0 release candidate 1 available

[Peter Bex]

On Tue, Jun 09, 2015 at 09:59:27PM +0200, Peter Bex wrote:
> Operating system: Debian 8.1
> Hardware platform: Powerpc (32 bits, iBook G4)
> C Compiler: Clang 3.5.0-10
> Installation works?: yes
> Tests work?: no
> Installation of eggs works?: no
> 
> The clang build segfaults on the very first test, and if I build
> with DEBUGBUILD=1 it fails with a panic:
> 
> [panic] `##sys#error-hook' is not defined - the `library' unit was probably 
> not linked with this executable - execution terminated
> 
> I'll try to investigate why this is failing later.

I just tried, and clang produces binaries that segfault for
CHICKEN 4.8.0 and CHICKEN 4.9.0 as well.  It had a very tough
time compiling CHICKEN 4.6.0 and 4.7.0; after an hour of waiting
for it to compile library.c, I decided to abort the build.

Because this isn't a new problem, I don't think we should let
this hold up the 4.10.0 release.  Considering clang's problems
with compiling even library.c for those older releases, I think
it's safe to assume (for now) that clang has problems on powerpc.

I tried digging in, but I hit a wall as soon as I tried gdb on a
basic csi session: it fails right at the start, in C_toplevel:
it looks like the toplevel_trampoline argument it passes to
C_reclaim (a function pointer) already gets mangled.  This might
be a problem with casting function pointers to C_word *.

Cheers.
Peter

signature.asc
Description: Digital signature