> Am 26.01.2016 um 06:37 schrieb David Gibson <da...@gibson.dropbear.id.au>:
>
>> On Mon, Jan 25, 2016 at 02:38:15PM +0000, Peter Maydell wrote:
>>> On 25 January 2016 at 14:00, David Gibson <da...@gibson.dropbear.id.au>
>>> wrote:
>>>> On Mon, Jan 25, 2016 at 11:59:22AM +0000, Peter Maydell wrote:
>>>>> On 25 January 2016 at 11:19, David Gibson <da...@gibson.dropbear.id.au>
>>>>> wrote:
>>>>> I think I've fixed it, but it's a bit hard to tell since for me
>>>>> origin/master also fails to compile on 32-bit :(.
>>>>
>>>> Oops -- what's the failure there?
>>>>
>>>> (I do know of a vixl compile failure with older gcc you might be running
>>>> into; if that's it try configuring with --cxx=none. I should send a patch
>>>> to fix that.)
>>>
>>> Doesn't look like it. Might be something to do with a specific glib
>>> version though:
>>>
>>> $ make -k
>>> CC qga/commands-posix.o
>>> qga/commands-posix.c: In function ‘qmp_guest_file_write’:
>>> qga/commands-posix.c:529:39: error: passing argument 3 of ‘qbase64_decode’
>>> from incompatible pointer type [-Werror=incompatible-pointer-types]
>>> buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
>>> ^
>>> In file included from qga/commands-posix.c:32:0:
>>> /home/dwg/src/qemu/include/qemu/base64.h:52:10: note: expected ‘size_t *
>>> {aka unsigned int *}’ but argument is of type ‘gsize * {aka long unsigned
>>> int *}’
>>> uint8_t *qbase64_decode(const char *input,
>>> ^
>>> cc1: all warnings being treated as errors
>>> /home/dwg/src/qemu/rules.mak:57: recipe for target 'qga/commands-posix.o'
>>> failed
>>> make: *** [qga/commands-posix.o] Error 1
>>> CC util/base64.o
>>
>> Ah, that's your compile environment being wrong -- you're trying to
>> build QEMU 32-bit but using the 64-bit glib headers.
>
> There are different versions of the *headers*!? Wtf?
>
>> Daniel sent a patch which diagnoses this at the configure stage
>> http://patchwork.ozlabs.org/patch/544254/
>> but it looks like it didn't get applied...
>>
>> Try setting PKG_CONFIG_LIBDIR and perhaps other things.
>
> Good grief. And this would be why I don't generally test 32-bit
> builds...
Just set up a 32bit vm and maybe configure it to automatically test your git
branch? ;)
Alex
