Re: [Qemu-devel] [PATCH] configure: correctly define PTHREAD_LIB

[Alexey Perevalov <a.perevalov@samsung.com>]

On 09/29/2017 04:05 PM, Daniel P. Berrange wrote:
> On Fri, Sep 29, 2017 at 03:52:34PM +0300, Alexey Perevalov wrote:
>> On 09/29/2017 01:40 PM, Daniel P. Berrange wrote:
>>> On Fri, Sep 29, 2017 at 01:11:14PM +0300, Alexey Perevalov wrote:
>>>> In case of -pthread already exists in QEMU_CFLAGS,
>>>> compilation of sample pthread program successed,
>>>> but -pthread is not putting into PTHREAD_LIBS in this case.
>>>> PTHREAD_LIB is using while compiling tests/migration/stress.
>>>>
>>>> Signed-off-by: Alexey Perevalov <a.perevalov@samsung.com>
>>>> ---
>>>>    configure | 1 +
>>>>    1 file changed, 1 insertion(+)
>>>>
>>>> diff --git a/configure b/configure
>>>> index 6587e80..a2dd044 100755
>>>> --- a/configure
>>>> +++ b/configure
>>>> @@ -3359,6 +3359,7 @@ int main(void) {
>>>>    EOF
>>>>    if compile_prog "" "" ; then
>>>>      pthread=yes
>>>> +  PTHREAD_LIB="-pthread"
>>>>    else
>>>>      for pthread_lib in $PTHREADLIBS_LIST; do
>>>>        if compile_prog "" "$pthread_lib" ; then
>>> We shouldn't do this because it affects whole of QEMU. The stress program
>>> needs -lpthread because it is linking statically, so just add it to the
>>> Makefile rule for building stress.
>> ok, but I didn't find any explicit usage of PTHREAD_LIB,
>> and avoiding it in Makefile I think will make PTHREAD_LIB
>> redundant.
> Oh yeah, actually PTHREAD_LIB is a variable i introduced specifically
> for the stress program, so my comment above is wrong and I think your
> patch is ok.
in that hardcoded form or trying to generalize it somehow?

-pthread comes to QEMU_CFLAGS from pkg-config --cflags for glib library
so it's hard to say which one from list -pthread -lpthread -lpthreadGC2
really lead to success build of the sample.
So it's better to cut pthread from QEMU_CFLAGS, or you thing it's overkill?

>
>
> Regards,
> Daniel


-- 
Best regards,
Alexey Perevalov