This is the mail archive of the libc-help@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: global variables within library __thread


Hi

compile like this works - ags_thread-posix.h:

static __thread AgsThread *ags_thread_self;

and within source file - ags_thread-posix.c:

extern __thread AgsThread *ags_thread_self;

but is it correct?

cheers,
JoÃl

On Sat, Oct 17, 2015 at 5:58 PM, JoÃl KrÃhemann <weedlight@gmail.com> wrote:
> Hi
>
> Putting into header has same effect as static global, isn't it?
> static global is as far as I now how you do a semaphore.
>
>
> On Sat, Oct 17, 2015 at 5:56 PM, JoÃl KrÃhemann <weedlight@gmail.com> wrote:
>> ... and as told these variables should be per process so putting into header
>> is definitively no option, think so.
>>
>>
>> On Sat, Oct 17, 2015 at 5:49 PM, JoÃl KrÃhemann <weedlight@gmail.com> wrote:
>>> Hi
>>>
>>> First of all it is located in the ags_thread-posix.c file but as
>>> putting it into the header i get
>>> something like:
>>>
>>> /usr/bin/ld: ags_thread_self: TLS definition in gsequencer-main.o
>>> section .tbss mismatches non-TLS reference in .libs/gsequencerS.o
>>> .libs/gsequencerS.o: error adding symbols: Bad value
>>> collect2: error: ld returned 1 exit status
>>> libtool: link: rm -f ".libs/gsequencerS.o"
>>> Makefile:2010: recipe for target 'gsequencer' failed
>>> make[1]: *** [gsequencer] Error 1
>>> make[1]: Leaving directory '/home/joelkraehemann/gsequencer'
>>> Makefile:1854: recipe for target 'all' failed
>>> make: *** [all] Error 2
>>>
>>> This reference isn't used anywhere but in ags_thread-posix.c. Further
>>> main.c doesn't contain any
>>> references to it. For completion here's the code of the functions using it:
>>>
>>>
>>> void
>>> ags_thread_resume_handler(int sig)
>>> {
>>>   if(ags_thread_self == NULL){
>>>     return;
>>>   }
>>>
>>> #ifdef AGS_DEBUG
>>>   g_message("thread resume\0");
>>> #endif
>>>
>>>   g_atomic_int_and(&(ags_thread_self->flags),
>>>                    (~AGS_THREAD_SUSPENDED));
>>>
>>>   ags_thread_resume(ags_thread_self);
>>> }
>>>
>>> void
>>> ags_thread_suspend_handler(int sig)
>>> {
>>> #ifdef AGS_DEBUG
>>>   g_message("thread suspend\0");
>>> #endif
>>>
>>>   if(ags_thread_self == NULL){
>>>     return;
>>>   }
>>>
>>>   if ((AGS_THREAD_SUSPENDED &
>>> (g_atomic_int_get(&(ags_thread_self->flags)))) != 0) return;
>>>
>>>   g_atomic_int_or(&(ags_thread_self->flags),
>>>                   AGS_THREAD_SUSPENDED);
>>>
>>>   ags_thread_suspend(ags_thread_self);
>>>
>>>   do sigsuspend(&(ags_thread_self->wait_mask)); while
>>> ((AGS_THREAD_SUSPENDED &
>>> (g_atomic_int_get(&(ags_thread_self->flags)))) != 0);
>>> }
>>>
>>>
>>> cheers,
>>> JoÃl
>>>
>>>
>>> On Sat, Oct 17, 2015 at 4:51 AM, Mike Frysinger <vapier@gentoo.org> wrote:
>>>> On 17 Oct 2015 00:31, JoÃl KrÃhemann wrote:
>>>>> __thread AgsThread *ags_thread_self = NULL;
>>>>>
>>>>> /usr/bin/ld: ags_thread_self: TLS definition in
>>>>> /home/joelkraehemann/gsequencer/.libs/libags_thread.a(libags_thread_la-ags_thread-posix.o)
>>>>> section .tbss mismatches non-TLS reference in .libs/gsequencerS.o
>>>>
>>>> make sure the header that declares this also has __thread markings.
>>>> declaring it in the header like this would be wrong:
>>>>         extern AgsThread *ags_thread_self;
>>>> -mike


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]