[Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, and gcc 11.0.0 20210202: build issues

Subramoni, Hari subramoni.1 at osu.edu
Mon Feb 8 17:54:56 EST 2021


Adding the correct e-mail address to MVAPICH-Discuss.

-----Original Message-----
From: Subramoni, Hari <subramoni.1 at osu.edu> 
Sent: Monday, February 8, 2021 5:15 PM
To: Jorge D'Elia <jdelia at cimec.unl.edu.ar>
Cc: mvapich-discuss at cse.ohio-state.edu <mvapich-discuss at mailman.cse.ohio-state.edu>; Subramoni, Hari <subramoni.1 at osu.edu>
Subject: RE: [Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, and gcc 11.0.0 20210202: build issues

Dear, Jorge.

It looks like the sys_siglist array has been deprecated in glib 2.3.2 and higher. This is probably why you're seeing the issue. Can you please try the following patch and see if it helps?

diff --git a/src/mpid/ch3/channels/common/src/util/error_handling.c b/src/mpid/ch3/channels/common/src/util/error_handling.c
index b8841bd..20a47ca 100644
--- a/src/mpid/ch3/channels/common/src/util/error_handling.c
+++ b/src/mpid/ch3/channels/common/src/util/error_handling.c
@@ -69,7 +69,7 @@ int show_backtrace = 0;  // Signal handler for errors  void error_sighandler(int sig, siginfo_t *info, void *secret) {
     // Always print error
-    PRINT_ERROR( "Caught error: %s (signal %d)\n", sys_siglist[sig], sig );
+    PRINT_ERROR( "Caught error: %s (signal %d)\n", strsignal(sig), sig 
+ );
     // Show backtrace if required
     if (show_backtrace) print_backtrace();
     // Raise the signal again with default handler

diff --git a/src/pm/mpirun/mpispawn.c b/src/pm/mpirun/mpispawn.c index 7123958..4c6a262 100644
--- a/src/pm/mpirun/mpispawn.c
+++ b/src/pm/mpirun/mpispawn.c
@@ -802,7 +802,7 @@ void child_handler(int signal)
     gethostname(my_host_name, MAX_HOST_LEN);

     rank = mt_id;
-    PRINT_DEBUG(DEBUG_Fork_verbose, "mpispawn child_handler: got signal %d: %s\n", signal, sys_siglist[signal]);
+    PRINT_DEBUG(DEBUG_Fork_verbose, "mpispawn child_handler: got signal 
+ %d: %s\n", signal, strsignal(signal));
     while (1) {
         do {
             pid = waitpid(-1, &status, WNOHANG); Best, Hari.

-----Original Message-----
From: Jorge D'Elia <jdelia at intec.unl.edu.ar>
Sent: Monday, February 8, 2021 5:02 PM
To: Subramoni, Hari <subramoni.1 at osu.edu>
Cc: mvapich-discuss at cse.ohio-state.edu <mvapich-discuss at mailman.cse.ohio-state.edu>
Subject: Re: [Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, and gcc 11.0.0 20210202: build issues

Dear Hari,

----- Mensaje original -----
> De: "Subramoni, Hari" 
> Para: "Jorge D'Elia" <jdelia at intec.unl.edu.ar>, "jdelia" 
> <jdelia at cimec.unl.edu.ar>
> CC: "mvapich-discuss at cse.ohio-state.edu" <mvapich-discuss at mailman.cse.ohio-state.edu>, "Subramoni, Hari"
> 
> Enviados: Lunes, 8 de Febrero 2021 16:38:02
> Asunto: RE: [Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, 
> and gcc 11.0.0 20210202: build issues
>
> Forgot to CC discuss in my original reply.
> 
> Hari.
> 
> -----Original Message-----
> From: Subramoni, Hari
> Sent: Monday, February 8, 2021 2:37 PM
> To: Jorge D'Elia <jdelia at intec.unl.edu.ar>; Jorge D'Elia 
> <jdelia at cimec.unl.edu.ar>
> Cc: Subramoni, Hari
> Subject: RE: [Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, 
> and gcc
> 11.0.0 20210202: build issues
> 
> Dear, Jorge.
> 
> Thanks for the report. We appreciate it.

Thanks for your fast reply!

> When we released MVAPICH2 2.3.5, the latest GCC available to us was 10.2.

Ok. Then, I tried with the GNU Fortran (GCC) 10.2.1 20201125 that is included in the Fedora 33 distribution. Nevertheless, now I have other errors in the make phase, please see the attached files.

> Redhat had reported a similar issue to us a few days ago. We are 
> working on setting up a local system with GCC11 to resolve this. We 
> will do this and get back to you as soon as possible.

Ok. Thanks!

Regards,
Jorge.

> -----Original Message-----
> From: Mvapich-discuss
> <mvapich-discuss-bounces+subramoni.1=osu.edu at lists.osu.edu> On Behalf 
> Of Jorge D'Elia via Mvapich-discuss
> Sent: Monday, February 8, 2021 1:41 PM
> To: mvapich-discuss at lists.osu.edu
> Subject: [Mvapich-discuss] MVAPICH2 2.3.5, Fedora Core 33 x86_64, and 
> gcc 11.0.0
> 20210202: build issues
> 
> Dear all,
> 
> I have some problems building the mvapich 2.3.5 on a laptop with 
> Fedora 33/x86_64.
> The common error message is:
> 
>  error: size of array element is not a multiple of its alignment
> 
> Below I paste some information and I also attach some relevant files.
> Please, some clue to solve it.
> 
> Thanks in advance.
> 
> Regards
> Jorge.
> 
> --
> #begin
> Build and Installation issues
> 
>    MVAPICH2 version:
>       MVAPICH2 2.3.5
>       Last revised: November 25, 2020
>       
>    Compiler version
>       gcc --version
>       gcc (GCC) 11.0.0 20210202 (experimental)
> 
>    Platform details (OS, kernel version..etc)
>       Fedora Core 33 x86_64
>       kernel-core-5.10.13-200.fc33.x86_64
>       
>    Configure flags
>       ../configure --with-device=ch3:mrail --with-rdma=gen2 --enable-fortran=all
>       --prefix=/usr/beta/mvapich2
>       
>    Attach config.log file
>        attached
> 
>    Attach configure, make and make install step output
>        ./configure {–flags} 2>&1 | tee config.out  --> attached
>        Make 2>&1 | tee make.out                    --> attached
>        Make install 2>&1 | tee install.out         --> not obtained
>    System information to include
>        OS version (/etc/issue)
>        Kernel version (uname -a)
>          Linux asgard 5.10.13-200.fc33.x86_64 #1 SMP
>          Thu Feb 4 14:54:51 UTC 2021 x86_64 x86_64 x86_64 GNU/Linux
>        OFED info (ofed_info)
>        IB card info (ibv_devinfo)
> #end


More information about the Mvapich-discuss mailing list