fix #30653, call malloc_trim sometimes (#32428)

this works around what seems to be a glibc bug
(cherry picked from commit f77743c)

src/gc.c

@@ -3,6 +3,9 @@
 #include "gc.h"
 #include "julia_gcext.h"
 #include "julia_assert.h"
+#ifdef __GLIBC__
+#include <malloc.h> // for malloc_trim
+#endif
 
 #ifdef __cplusplus
 extern "C" {
@@ -555,6 +558,10 @@ static int prev_sweep_full = 1;
 // Full collection heuristics
 static int64_t live_bytes = 0;
 static int64_t promoted_bytes = 0;
+#ifdef __GLIBC__
+// maxrss at last malloc_trim
+static int64_t last_trim_maxrss = 0;
+#endif
 
 static int64_t last_full_live_ub = 0;
 static int64_t last_full_live_est = 0;
@@ -2685,6 +2692,8 @@ static void jl_gc_queue_bt_buf(jl_gc_mark_cache_t *gc_cache, jl_gc_mark_sp_t *sp
  }
 }
 
+size_t jl_maxrss(void);
+
 // Only one thread should be running in this function
 static int _jl_gc_collect(jl_ptls_t ptls, int full)
 {
@@ -2841,6 +2850,18 @@ static int _jl_gc_collect(jl_ptls_t ptls, int full)
  }
  }
 
+#ifdef __GLIBC__
+ if (sweep_full) {
+ // issue #30653
+ // empirically, the malloc runaway seemed to occur within a growth gap
+ // of about 20-25%
+ if (jl_maxrss() > (last_trim_maxrss/4)*5) {
+ malloc_trim(0);
+ last_trim_maxrss = jl_maxrss();
+ }
+ }
+#endif
+
  uint64_t gc_end_t = jl_hrtime();
  uint64_t pause = gc_end_t - t0;
  gc_final_pause_end(t0, gc_end_t);