Re: [Xen-devel] [PATCH] tools: adjust --datadir option passed to qemu-upstream's configure script

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 08.06.12 at 13:11, Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
wrote:
> On Mon, 4 Jun 2012, Jan Beulich wrote:
>> Passing $(SHAREDIR)/qemu-xen isn't compatible with the way
>> os-posix.c:os_find_datadir() works. While one would generally expect
>> that function to honour the --datadir configure option, fixing this is
>> quite a bit more involved than simply adjusting the configure options
>> to match the function's behavior (in that, for an installed binary, it
>> looks in $bindir/../share/qemu). Afaict, so far this can have worked
>> only by virtue of what CONFIG_QEMU_DATADIR points to being populated
>> due to (presumably) some other qemu instance being installed on
>> people's systems (but being absent when running qemu-system-i386 from
>> the dist/ portion of the build tree).
>> 
>> Signed-off-by: Jan Beulich <JBeulich@xxxxxxxx>
>> 
>> --- a/tools/Makefile
>> +++ b/tools/Makefile
>> @@ -155,7 +155,7 @@ subdir-all-qemu-xen-dir subdir-install-q
>>              --extra-ldflags="-L$(XEN_ROOT)/tools/libxc \
>>              -L$(XEN_ROOT)/tools/xenstore" \
>>              --bindir=$(LIBEXEC) \
>> -            --datadir=$(SHAREDIR)/qemu-xen \
>> +            --datadir=$(dir $(LIBEXEC))share/qemu \
>>              --disable-kvm \
>>              --python=$(PYTHON) \
>>              $(IOEMU_CONFIGURE_CROSS); \
>> 
>> 
>> 
>> 
> 
> This is a QEMU bug, so I would rather fix it there.

That's not as simple as you appear to think, as it would be necessary
to meaningfully express an arbitrary datadir relative to bindir.

> In fact the fix could be as simple as the appended patch.
> Does it work for you?
> 
> diff --git a/os-posix.c b/os-posix.c
> index dc4a6bb..d0c873d 100644
> --- a/os-posix.c
> +++ b/os-posix.c
> @@ -136,6 +136,9 @@ char *os_find_datadir(const char *argv0)
>              g_free(res);
>              res = NULL;
>          }
> +    } else {
> +        g_free(res);
> +        res = NULL;

How can this work? Afaict it would even break currently working
cases, as it now makes the function return NULL not only when
both access() checks failed, but also when the first one succeeded.

Jan

>      }
>  
>      return res;




_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel