From 6963302bb58ba0b4530c9503f8b26bb71c691fb5 Mon Sep 17 00:00:00 2001
From: Augustin Degomme <adegomme@users.noreply.github.com>
Date: Mon, 27 Sep 2021 00:24:00 +0200
Subject: [PATCH] Intercept exit() to avoid killing simulations too early, as
 the first process exits. This should render previous warning pointless, let's
 see it it's still needed.

---
 ChangeLog                                   | 3 +++
 include/smpi/smpi_helpers.h                 | 2 ++
 include/smpi/smpi_helpers_internal.h        | 1 +
 src/smpi/internals/smpi_global.cpp          | 6 ++++++
 teshsuite/smpi/pt2pt-dsend/pt2pt-dsend.tesh | 2 --
 5 files changed, 12 insertions(+), 2 deletions(-)

diff --git a/ChangeLog b/ChangeLog
index 13e0736676..5b778314bb 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -28,6 +28,9 @@ SMPI:
    - scan/excan can now be replayed
    - wait action now uses ranks and not pid, as the other ones.
    - smpi/init and smpi/finalization-barrier are now valid for replays.
+ - exit() is now intercepted by SMPI to avoid premature shutdown of
+   simulation. First non 0 return codes is returned as simulation return
+   code.
 
 Documentation:
   * New section "Release Notes" documenting recent and current developments.
diff --git a/include/smpi/smpi_helpers.h b/include/smpi/smpi_helpers.h
index ef7492d998..8ecd7f582a 100644
--- a/include/smpi/smpi_helpers.h
+++ b/include/smpi/smpi_helpers.h
@@ -38,4 +38,6 @@
 #define realloc(x, y) smpi_shared_realloc_intercept((x), (y), __FILE__, __LINE__)
 #define free(x) smpi_shared_free(x)
 #endif
+
+#define exit(x) smpi_exit(x)
 #endif
diff --git a/include/smpi/smpi_helpers_internal.h b/include/smpi/smpi_helpers_internal.h
index 25a4f7093b..929d22e75e 100644
--- a/include/smpi/smpi_helpers_internal.h
+++ b/include/smpi/smpi_helpers_internal.h
@@ -38,6 +38,7 @@ void* smpi_shared_malloc_intercept(size_t size, const char* file, int line);
 void* smpi_shared_calloc_intercept(size_t num_elm, size_t elem_size, const char* file, int line);
 void* smpi_shared_realloc_intercept(void* data, size_t size, const char* file, int line);
 void smpi_shared_free(void* data);
+void smpi_exit(int status);
 #ifdef __cplusplus
 } // extern "C"
 #endif
diff --git a/src/smpi/internals/smpi_global.cpp b/src/smpi/internals/smpi_global.cpp
index 674cfed28c..13c6a3be99 100644
--- a/src/smpi/internals/smpi_global.cpp
+++ b/src/smpi/internals/smpi_global.cpp
@@ -635,3 +635,9 @@ void SMPI_thread_create() {
   TRACE_smpi_init(simgrid::s4u::this_actor::get_pid(), __func__);
   smpi_process()->mark_as_initialized();
 }
+
+void smpi_exit(int x){
+  if(x!=0)
+    smpi_exit_status = x;
+  return;
+}
diff --git a/teshsuite/smpi/pt2pt-dsend/pt2pt-dsend.tesh b/teshsuite/smpi/pt2pt-dsend/pt2pt-dsend.tesh
index 713e5410bc..11476da66f 100644
--- a/teshsuite/smpi/pt2pt-dsend/pt2pt-dsend.tesh
+++ b/teshsuite/smpi/pt2pt-dsend/pt2pt-dsend.tesh
@@ -3,7 +3,6 @@ p Test dsend
 $ ${bindir:=.}/../../../smpi_script/bin/smpirun -map -hostfile ${bindir:=.}/../hostfile -platform ${platfdir}/small_platform.xml -np 2 --log=no_loc ${bindir:=.}/pt2pt-dsend -s --long --log=smpi_config.thres:warning --log=xbt_cfg.thres:warning --cfg=smpi/simulate-computation:no --cfg=smpi/finalization-barrier:on
 > [Jupiter:1:(2) 0.000000] [dsend/INFO] rank 1: data exchanged
 > [Tremblay:0:(1) 0.005896] [dsend/INFO] rank 0: data exchanged
-> [Tremblay:0:(1) 0.008843] [ker_engine/WARNING] Process called exit when leaving - Skipping cleanups
 > [0.000000] [smpi/INFO] [rank 0] -> Tremblay
 > [0.000000] [smpi/INFO] [rank 1] -> Jupiter
 
@@ -14,7 +13,6 @@ p process 2 will finish at 0.5+2*4 (time before first send) + 2*(1+0.5*4) (recv+
 ! output sort
 $ ${bindir:=.}/../../../smpi_script/bin/smpirun -map -hostfile ${bindir:=.}/../hostfile -platform ${platfdir}/small_platform.xml -np 2 --log=no_loc ${bindir:=.}/pt2pt-dsend -s --long --log=smpi_config.thres:warning --cfg=smpi/or:0:1:0.5 --cfg=smpi/os:0:0.5:2 --cfg=smpi/ois:0:1:0.1 --cfg=smpi/simulate-computation:no --cfg=smpi/finalization-barrier:on --log=xbt_cfg.thres:warning
 > [Jupiter:1:(2) 9.900000] [dsend/INFO] rank 1: data exchanged
-> [Tremblay:0:(1) 16.508843] [ker_engine/WARNING] Process called exit when leaving - Skipping cleanups
 > [Tremblay:0:(1) 14.505896] [dsend/INFO] rank 0: data exchanged
 > [0.000000] [smpi/INFO] [rank 0] -> Tremblay
 > [0.000000] [smpi/INFO] [rank 1] -> Jupiter
-- 
2.20.1