On 6 February 2014 23:38, Paolo Bonzini <pbonz...@redhat.com> wrote:
>> Looks like we're not getting the executable path
>> correctly on MacOS for some reason.
>
> Indeed, there is no code for it. Can you try the patch below?
>
>> Also, mjt pointed out on IRC that we probably want to allow
>> installing binary modules into a path with the arch name in
>> it, to allow for multiarch distros.
>
> Isn't --libdir enough?
>
> Paolo
>
> ==================== 8< ========================
> From: Paolo Bonzini <pbonz...@redhat.com>
> Date: Wed, 5 Feb 2014 21:15:01 +0100
> Subject: [PATCH] oslib: port qemu_exec_dir to Darwin
>
> Signed-off-by: Paolo Bonzini <pbonz...@redhat.com>
> ---
> util/oslib-posix.c | 15 +++++++++++++++
> 1 file changed, 15 insertions(+)
>
> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
> index 372b2f9..590a7d1 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -59,6 +59,12 @@ extern int daemon(int, int);
> #include <sys/mman.h>
> #include <libgen.h>
>
> +/* Get declaration of _NSGetExecutablePath on MacOS X 10.2 or newer. */
> +#if defined(__APPLE__) && defined(__MACH__)
> +#define ENUM_DYLD_BOOL
> +#include <mach-o/dyld.h>
> +#endif
> +
> #ifdef CONFIG_LINUX
> #include <sys/syscall.h>
> #endif
> @@ -303,6 +309,15 @@ char *qemu_exec_dir(const char *argv0)
> p = buf;
> }
> }
> +#elif defined(__APPLE__) && defined(__MACH__)
> + {
> + char result[PATH_MAX];
> + uint32_t length = PATH_MAX;
> + if (_NSGetExecutablePath (result, &length) != 0 || result[0] != '/')
> {
> + return NULL;
> + }
> + p = realpath(result, buf);
> + }
> #endif
> /* If we don't have any way of figuring out the actual executable
> location then try argv[0]. */
> --
> 1.8.5.3
Why do we need OS specific code to do this when the code
in vl.c for getting the executable path to use as a base for
finding bios blobs works fine without OS specific code?
thanks
-- PMM
