fix #8551 and make stack switch more efficient

previously, each new stack would have a copy of a small portion of the
stack that started it. in small amounts, this was fine, but it is bad if
the user is starting many tasks. additionally, it wasted memory inside
every task to have a per-task base_ctx.

(cherry picked from commit ac13711)

src/julia.h

@@ -992,11 +992,12 @@ DLLEXPORT void jl_handle_stack_switch();
 
 #ifdef COPY_STACKS
 // initialize base context of root task
+extern jl_jmp_buf jl_base_ctx;
 #define JL_SET_STACK_BASE                               \
     {                                                   \
         int __stk;                                      \
         jl_root_task->stackbase = (char*)&__stk;        \
-        if (jl_setjmp(jl_root_task->base_ctx, 1)) {     \
+        if (jl_setjmp(jl_base_ctx, 1)) {                \
             jl_handle_stack_switch();                   \
         }                                               \
     }
@@ -1145,7 +1146,6 @@ typedef struct _jl_task_t {
         void *stackbase;
         void *stack;
     };
-    jl_jmp_buf base_ctx;
     size_t bufsz;
     void *stkbuf;
     size_t ssize;

src/task.c

@@ -30,6 +30,8 @@ volatile int jl_in_stackwalk = 0;
 #include <dlfcn.h>   // for dladdr
 #endif
 
+jl_jmp_buf jl_base_ctx;
+
 /* This probing code is derived from Douglas Jones' user thread library */
 
 /* true if stack grows up, false if down */
@@ -211,11 +213,6 @@ static void switch_stack(jl_task_t *t, jl_jmp_buf *where)
         restore_stack(t, where, NULL);
     }
 }
-
-void jl_switch_stack(jl_task_t *t, jl_jmp_buf *where)
-{
-    switch_stack(t, where);
-}
 #endif
 
 static void ctx_switch(jl_task_t *t, jl_jmp_buf *where)
@@ -262,7 +259,7 @@ static void ctx_switch(jl_task_t *t, jl_jmp_buf *where)
 
 #ifdef COPY_STACKS
         jl_jmp_target = where;
-        jl_longjmp(lastt->base_ctx, 1);
+        jl_longjmp(jl_base_ctx, 1);
 #else
         jl_longjmp(*where, 1);
 #endif
@@ -418,10 +415,6 @@ static void start_task(jl_task_t *t)
     local_sp += sizeof(jl_gcframe_t);
     local_sp += 12*sizeof(void*);
     t->stackbase = (void*)(local_sp + _frame_offset);
-    if (jl_setjmp(t->base_ctx, 0)) {
-        // we get here to remove our data from the process stack
-        switch_stack(jl_current_task, jl_jmp_target);
-    }
 #endif
     res = jl_apply(t->start, NULL, 0);
     JL_GC_POP();
@@ -431,7 +424,7 @@ static void start_task(jl_task_t *t)
 
 DLLEXPORT void jl_handle_stack_switch()
 {
-    jl_switch_stack(jl_current_task, jl_jmp_target);
+    switch_stack(jl_current_task, jl_jmp_target);
 }
 
 #ifndef COPY_STACKS