How to disable "noreturn" attribute propagation/handling?

2017-04-21 Thread Thomas Schwinge 
Hi!

On Tue, 20 May 2014 13:21:06 +0200, Tom de Vries  wrote:
> On 17/05/14 17:53, Tom de Vries wrote:
> > On 16/05/14 22:28, Thomas Schwinge wrote:
> >> > It may make sense to have an abort function available in PTX land.

... which we've now had for a long time.  However, I recently noticed
that it doesn't actually behave as expected:
 'The nvptx
"abort" implementation doesn't actually abort execution #5'.  ;-| That
is, for the last year or so, the nvptx "abort" implementation silently
"exit"ed instead of "trap"ping.  However code such as:

> #pragma acc parallel
>   {
> abort ();
>   }

... still caused process termination, and while I have not analyzed that
in detail, the reason seems to be along the lines that GCC propagates the
"abort"'s "noreturn" attribute to the OMP child function
("main._omp_fn.0"), and then places a "__builtin_unreachable" after the
"GOACC_parallel_keyed" -- which is the cause of the process termination.
(Quite surprising to me, that is done even for "-O0"!)  See
abort-1.c.019t.fixup_cfg1:

[...]
Introduced new external node (__builtin_unreachable/8).
[...]
  GOACC_parallel_keyed (-1, main._omp_fn.0, 0, 0B, 0B, 0B, 0);
  __builtin_unreachable ();
[...]

Now, I could do something like:

 #pragma acc parallel
   {
-abort ();
+int cond = 1;
+asm volatile ("" : : : "memory");
+if (cond)
+  abort ();
   }

... to prevent that from happening -- but isn't there a generic way
(command-line flag?) to disable the analysis that causes the "noreturn"
attribute propagation/handling?


Grüße
 Thomas

Machine problems at gcc.gnu.org?

2017-04-21 Thread Steve Ellcey 

I am having problems getting to https://gcc.gnu.org this morning and
I have also had problems getting to the glibc mail archives though the
main web page for glibc seem available.  Anyone else having problems?
Of course if this email goes through the machines that are having problems
it may not get anywhere

Steve Ellcey
sell...@cavium.com

Re: Machine problems at gcc.gnu.org?

2017-04-21 Thread Markus Trippelsdorf 
On 2017.04.21 at 09:17 -0700, Steve Ellcey wrote:
> 
> I am having problems getting to https://gcc.gnu.org this morning and
> I have also had problems getting to the glibc mail archives though the
> main web page for glibc seem available.  Anyone else having problems?
> Of course if this email goes through the machines that are having problems
> it may not get anywhere

Yes, looks like the sourceware server is having problems:
https://check-host.net/check-ping?host=gcc.gnu.org

-- 
Markus

Re: Machine problems at gcc.gnu.org?

2017-04-21 Thread H.J. Lu 
On Fri, Apr 21, 2017 at 9:17 AM, Steve Ellcey  wrote:
>
> I am having problems getting to https://gcc.gnu.org this morning and
> I have also had problems getting to the glibc mail archives though the
> main web page for glibc seem available.  Anyone else having problems?
> Of course if this email goes through the machines that are having problems
> it may not get anywhere
>

Yes, I have the same problem.


-- 
H.J.

Re: Machine problems at gcc.gnu.org?

2017-04-21 Thread David Edelsohn 
gcc.gnu.org and sourceware.org are hosted by Red Hat datacenter, which
apparently experienced a widespread problem.

http://status.redhat.com/

On Fri, Apr 21, 2017 at 12:17 PM, Steve Ellcey  wrote:
>
> I am having problems getting to https://gcc.gnu.org this morning and
> I have also had problems getting to the glibc mail archives though the
> main web page for glibc seem available.  Anyone else having problems?
> Of course if this email goes through the machines that are having problems
> it may not get anywhere
>
> Steve Ellcey
> sell...@cavium.com