static symbol being made globally visible: compiler problem?

That would be a bug.  I've submitted SDSCM00048452 to track this issue.  As a workaround, give the static functions different names.

Thanks. Yes, I had already implemented that workaround; I just wanted to make sure that I wasn't missing something about "static inline".

SPH

based on your description of the problem, I worked out the following testcase:

====test1.c

#include <stdio.h>

static inline ResetState(void)
{

printf("In ResetState in test1\n");
}

extern void GloballyUniqueName1(void)
{
ResetState();
}

====

====test2.c

#include <stdio.h>

static inline ResetState(void)
{
printf("In ResetState in test2\n");

}

/* Lots of other code where ResetState gets called and is correctly inlined */

extern void GloballyUniqueName2(void)
{
ResetState();
}

====

====test3.c

#include <stdio.h>

extern void GloballyUniqueName1();
extern void GloballyUniqueName2();

main()
{
GloballyUniqueName1();
GloballyUniqueName2();
}

And I use these command to compile them:

cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k test1.c
cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k test2.c
cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k test3.c
cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k test1.obj test2.obj test3.obj -z -o test.out -llnk.cmd"test1.c", line 3: remark: explicit type is missing ("int" assumed)

"test1.c",line 7: remark: missing return statement at end of non-void function "ResetState"

"test2.c", line 3: remark: explicit type is missing ("int" assumed)
"test2.c", line 7: remark: missing return statement at end of non-void function "ResetState"
"test3.c", line 10: remark: missing return statement at end of non-void function "main"
remark: automatic RTS selection: linking in index library "libc.a"
remark: automatic RTS selection: linking in "rts64plus_elf.lib" in place of
index library "libc.a"

The linking is done fine.

When I run it, I got the following:

In ResetState in test1
In ResetState in test2

It seems this is also correct.

Please revisit the testcase I wrote based on your description, maybe I missed something. Otherwise, it is OK. I did not see any problem.

Thanks,

Wei

OK, I've reproduced it:

static int a;

static inline void ResetState(void)
{
   a = 0;
}

extern void f1_a(void)
{
   volatile int local;
   int i;

   for(i = 0; i < 10; i++)
   {
      local = i;
   }
   ResetState();
}

extern void f1_b(void)
{
   ResetState();
}

static int a;

static inline void ResetState(void)
{
   a = 0;
}

extern void f2_a(void)
{
   volatile int local;
   int i;

   for(i = 0; i < 10; i++)
   {
      local = i;
   }
   ResetState();
}

extern void f2_b(void)
{
   ResetState();
}

extern void f1_a(void);
extern void f1_b(void);
extern void f2_a(void);
extern void f2_b(void);

int main(void)
{
   f1_a();
   f2_a();
   f1_b();
   f2_b();

   return 0;
}

C:\tmp\ti_comp_issue>cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k f1.c
"f1.c", line 1: warning: variable "a" was set but never used

C:\tmp\ti_comp_issue>cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k f2.c
"f2.c", line 1: warning: variable "a" was set but never used

C:\tmp\ti_comp_issue>cl6x -qq --gcc -O3 -mi200 -mv6455 -mt -mw -oi -pdr --abi=eabi -k main.c

C:\tmp\ti_comp_issue>lnk6x *.obj
error: symbol "ResetState" redefined: first defined in "f1.obj"; redefined in
   "f2.obj"
error: no input section is linked in
warning: no suitable entry-point found; setting to 0
error: errors encountered during linking; "a.out" not built

Both ResetState functions are, as you can see, defined as "static inline void ResetState(void)".

SPH

I've just had a bit more of a play and found that it looks to be a problem specifically with the EABI; I get the problem if I compile using "cl6x --abi=eabi -O1 <file>.c" but not if I compile with "cl6x -O1 <file>.c". All the other options appear superfluous to reproducing the problem.

Cheers,

SPH

Ah, OK. I've just tried with a newer version of the tools and can confirm that the problem has gone away.

Thanks for your help,

SPH

Great!

In the future, if possible, please always try to use the newest version of TI tools. It may solve some problems you know or you don't even feel.

Thanks again for filing this problem to us,

Wei