well-tested compiler flags essential to gain useful backtrace(3) result

[RushOnline]

without this patch call to backtrace(3) returns zero.

Tested on RPiZ without analog/digital gain control support in ThreadX API:

Output of modified raspistill -o /dev/null -awb off -awbg 1.56,1.78 -ex off -ag 1 -dg 1 -t 1 -v before and after patch:

-Obtained 0 stack frames.
+Obtained 7 stack frames.
+./raspistill(print_trace+0x1c) [0x11bcc]
+./raspistill(mmal_status_to_int+0x30) [0x12184]
+./raspistill(raspicamcontrol_set_gains+0x88) [0x1048c]
+./raspistill(raspicamcontrol_set_all_parameters+0x39c) [0xeef8]
+./raspistill() [0x1561c]
+./raspistill(main+0x228) [0x1798c]
+/lib/arm-linux-gnueabihf/libc.so.6(__libc_start_main+0x114) [0xb6c18678]
mmal: Function not implemented

raspistill modification (just for test, not included in this pull request):

diff --git a/host_applications/linux/apps/raspicam/RaspiHelpers.c b/host_applications/linux/apps/raspicam/RaspiHelpers.c
index 5c25e18..86e113e 100644
--- a/host_applications/linux/apps/raspicam/RaspiHelpers.c
+++ b/host_applications/linux/apps/raspicam/RaspiHelpers.c
@@ -33,6 +33,7 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 #include <unistd.h>
 #include <stdint.h>
 
+#include <execinfo.h>
 
 #include "interface/vcos/vcos.h"
 #include "interface/mmal/mmal.h"
@@ -58,6 +59,25 @@ SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
 static const char *app_name;
 
+/* Obtain a backtrace and print it to stdout. */
+void print_trace (void)
+{
+  void *array[10];
+  size_t size;
+  char **strings;
+  size_t i;
+
+  size = backtrace (array, 10);
+  strings = backtrace_symbols (array, size);
+
+  printf ("Obtained %zd stack frames.\n", size);
+
+  for (i = 0; i < size; i++)
+     printf ("%s\n", strings[i]);
+
+  free (strings);
+}
+
 void print_app_details(FILE *fd)
 {
    if (!app_name)
@@ -225,6 +245,7 @@ int mmal_status_to_int(MMAL_STATUS_T status)
       return 0;
    else
    {
+         print_trace();
       switch (status)
       {
       case MMAL_ENOMEM :

[popcornmix]

No real objections to this. @6by9 @pelwell ?

[pelwell]

Fine by me, provided there is little overhead.

[6by9]

It's for debug builds only, so I see little downside. I don't know whether it needs all the debug options (-g3 vs -g), but other than library/exe size it won't do any harm, likewise -O0.

[RushOnline]

And you don't need the -funwind-tables if you have debug information around somewhere (e.g. by adding "-g").

backtrace(3) without -funwind-tables returns zero for me.

[Ruffio]

@RushOnline has this been resolved?

[RushOnline]

@Ruffio I don't test it since I've posted this MR. May be fresh software don't have this bug.

[Ruffio]

It would be nice to get this closed, if the bug doesn't persists with fresh software