On Fri, Jan 18, 2013 at 8:36 AM, Paolo Bonzini <pbonz...@redhat.com> wrote:
> Il 17/01/2013 22:18, Blue Swirl ha scritto:
>> On Thu, Jan 17, 2013 at 8:54 PM, Stefan Weil <s...@weilnetz.de> wrote:
>>> Am 17.01.2013 21:45, schrieb Blue Swirl:
>>>
>>>> On Wed, Jan 16, 2013 at 6:04 PM, Stefan Weil<s...@weilnetz.de> wrote:
>>>>>
>>>>> MinGW has no strtok_r, so we need a declaration in sysemu/os-win32.h.
>>>>> We must also fix the include statements in util/envlist.c to include
>>>>> that file.
>>>>>
>>>>> We currently don't need an implementation of strtok_r because the
>>>>> code is compiled but not linked for MinGW.
>>>>
>>>>
>>>> I think it would be better to fix the build system so that unnecessary
>>>> files are not compiled.
>>>
>>>
>>> That's what I suggested first, but keeping things simple is also
>>> a good argument. Perhaps we should accept that libqemuutil.a
>>> can contain some unnecessary files (that's the status quo!).
>>
>> This could be related. I get build errors on OpenBSD:
>> LINK i386-bsd-user/qemu-i386
>> ../libqemuutil.a(oslib-posix.o)(.text+0x540): In function `qemu_vmalloc':
>> /src/qemu/util/oslib-posix.c:112: multiple definition of `qemu_vmalloc'
>> bsd-user/mmap.o(.text+0x400):/src/qemu/bsd-user/mmap.c:78: first defined here
>> /usr/bin/ld: Warning: size of symbol `qemu_vmalloc' changed from 260
>> in bsd-user/mmap.o to 124 in ../libqemuutil.a(oslib-posix.o)
>
> Does this fix it?
>
> diff --git a/bsd-user/mmap.c b/bsd-user/mmap.c
> index 5d6cffc..1f4d4a3 100644
> --- a/bsd-user/mmap.c
> +++ b/bsd-user/mmap.c
> @@ -74,7 +74,7 @@ void mmap_unlock(void)
> }
> #endif
>
> -void *qemu_vmalloc(size_t size)
> +static void *qemu_vmalloc(size_t size)
> {
> void *p;
> mmap_lock();
No, it conflicts with the function prototype.
>
> (A better change is of course to also rename qemu_vmalloc).
This works.
>
>>>
>>> We also get the additional benefit of more portable code.
>>> Even if that portability is not needed for the moment,
>>> it might be useful later.
>>
>> Yes, though building as an example KVM code on Win32 may be useful one
>> day, that day may be distant.
>
> That's not the kind of extra code that Stefan is talking about. It's
> extra data structures (envlist.c) that are placed in a static library
> but do not appear in any final build product.
>
> Instead, "building KVM code on Win32" is in fact something we have been
> doing for years, see kvm-stub.c.
I think a similar case would be to compile the non-stub versions on
Win32 also but not link them in. For example, in theory envlist.c
could use POSIX only features.
>
> Paolo
