[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

Artem Polyakov  changed:

   What|Removed |Added

 Status|WAITING |RESOLVED
 Resolution|--- |INVALID

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #13 from Artem Polyakov  ---
Oh, now I see what was going on!

This is a really masked error:

while( (rc = connect(sd, (struct sockaddr*), addrlen) < 0) && (errno ==
EAGAIN) );

instead of:

while( (rc = connect(sd, (struct sockaddr*), addrlen)) < 0 && (errno ==
EAGAIN) );

so the problem is in the code, not GCC: wrong operation sequence:
rc = (connect(sd, (struct sockaddr*), addrlen) < 0)

instead of

(rc = (connect(sd, (struct sockaddr*), addrlen)) < 0


So we can safely close this bug report. Sorry for the false alarm.

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #12 from Artem Polyakov  ---
Created attachment 38610
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38610=edit
req.s_gcc-6.1.0

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #11 from Artem Polyakov  ---
Created attachment 38609
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38609=edit
req.i_gcc-6.1.0

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #10 from Artem Polyakov  ---
The same with gcc-6.1.0 (attached)

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #9 from Artem Polyakov  ---
Created attachment 38605
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38605=edit
req.s_gcc-5.3.0

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #7 from Artem Polyakov  ---
I checked with gcc-5.3.0 and I see the same results. Corresponding preprocessed
source and assembly files are attached.

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-31 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #8 from Artem Polyakov  ---
Created attachment 38604
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38604=edit
req.i_gcc-5.3.0

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-30 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #6 from Artem Polyakov  ---
BTW, the requested files are posted.

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-30 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #5 from Artem Polyakov  ---
Thank you, Marek.

I will check gcc 5.3.

However SLURM is mostly used on RHEL systems and it seems that RHEL is very
conservative about gcc:

RHEL 6.5: gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4)
RHEL 7.3: gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4)

So I'd like to have some guidance regarding those old versions. And find out
precise circumstances where this problem occurs.

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-30 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #3 from Artem Polyakov  ---
Created attachment 38602
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38602=edit
preprocessed source file

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-30 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #4 from Artem Polyakov  ---
Created attachment 38603
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=38603=edit
assembly file (req.s)

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-30 Thread mpolacek at gcc dot gnu.org
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

Marek Polacek  changed:

   What|Removed |Added

 Status|UNCONFIRMED |WAITING
   Last reconfirmed||2016-05-30
 CC||mpolacek at gcc dot gnu.org
 Ever confirmed|0   |1

--- Comment #2 from Marek Polacek  ---
We'll need a preprocessed source file, see https://gcc.gnu.org/bugs/.  Also GCC
4.8 is not supported anymore.

[Bug c/71287] Possibly broken invocation of connect (and maybe other) syscalls

2016-05-26 Thread artpol84 at gmail dot com
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71287

--- Comment #1 from Artem Polyakov  ---
On RHEL where this was also observed compiler version is the following:

gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-4)