Bug#574717: mingw32-ocaml: Supported arches
Le samedi 20 mars 2010 08:21:53, Kurt Roeckx a écrit :
> Hi,
Hi,
> It seems you currently restrict the arch list to:
> i386 amd64 kfreebsd-amd64 kfreebsd-i386
>
> Is there a reason for this? Both mingw32 and ocaml seem to be
> supported on all arches (except for mingw32 on hurd-i386).
>
> I also don't get why you need to build-depend on gcc-multilib.
This is needed to get a working build environment for gcc -m32
> You have this in your rules file:
> # Note that we must build a 32 bit compiler, even on 64 bit build
> # architectures, because this compiler will try to do strength
> # reduction optimizations using its internal int type, and that
> must # match Windows' int type. (That's what -cc and -host are for).
>
> But the size of an int doesn't change for 32 and 64 targets. Only the size
> of a long and a pointer do.
>
> The description even says:
> This compiler runs quickly,
> generates compact code with moderate memory requirements, and is
> portable to essentially any 32 or 64 bit Unix platform.
>
> I currently see no reason not to set the arch list to any.
The cross-compilation process is a bit more complicated than that. In
particular, the ocamlopt compiler emits temporary assembly files that are
compiled later by the compiler. When switching to gcc instead of gcc -m32, I
get a lot of messages like this:
/tmp/camlasmf9e413.s: Assembler messages:
/tmp/camlasmf9e413.s:431: Warning: value 0x3cb0000000000000 truncated to 0x0
/tmp/camlasmf9e413.s:437: Warning: value 0x10000000000000 truncated to 0x0
/tmp/camlasmf9e413.s:443: Warning: value 0x7fefffffffffffff truncated to 0xffffffff
/tmp/camlasmf9e413.s:449: Warning: value 0x7ff0000000000001 truncated to 0x1
/tmp/camlasmf9e413.s:455: Warning: value 0xfff0000000000000 truncated to 0x0
/tmp/camlasmf9e413.s:461: Warning: value 0x7ff0000000000000 truncated to 0x0
Hence, I assume it is not safe to compile it using (64 bits) gcc.
However, I'm open to any suggestion, in particular when tested and if possible
with a patch...
Romain
Reply to: