Bug #4630: PROCDEBUG to non-zero values causes segmentation faults in check_env in trunk

• Status: Closed
• Assignee: ko1 (Koichi Sasada)
• Target version: 2.6
• ruby -v: -
• Backport: 2.2: UNKNOWN, 2.3: UNKNOWN, 2.4: UNKNOWN

[ehuard (Elise Huard)]

PROCDEBUG is meant to show debugging output from the vm.
However, when it's activated in vm.c
#define PROCDEBUG 1
The make process fails in segmentation faults.

I've narrowed it down to the following variables displayed in check_env: env->block.dfp[3] and env->block.dfp[4]
Commenting them out like so makes the segfaults disappear.

  static int
  check_env(rb_env_t * const env)
  {
      printf("---\n");
      printf("envptr: %p\n", (void *)&env->block.dfp[0]);
      printf("orphan: %p\n", (void *)env->block.dfp[1]);
      printf("inheap: %p\n", (void *)env->block.dfp[2]);
      /* printf("envval: %10p ", (void *)env->block.dfp[3]);
      dp(env->block.dfp[3]);
      printf("penvv : %10p ", (void *)env->block.dfp[4]);
       dp(env->block.dfp[4]); */
      printf("lfp:    %10p\n", (void *)env->block.lfp);
      printf("dfp:    %10p\n", (void *)env->block.dfp);
    /*  if (env->block.dfp[4]) {
          printf(">>\n");
          check_env_value(env->block.dfp[4]);
          printf("<<\n");
      } */
      return 1;
  }

However, I'm not clued up enough on the inner workings of the VM to say why or how that should be fixed intelligently.

[ko1 (Koichi Sasada)]

• ruby -v changed from 1.9.3dev to -

Hi,

(2011/04/29 4:52), Elise Huard wrote:

PROCDEBUG is meant to show debugging output from the vm.
However, when it's activated in vm.c
#define PROCDEBUG 1
The make process fails in segmentation faults.

Oops. I don't touch it on long time (because there is no bug which
requires this debug option). So maybe this check_env() doesn't catch up
Proc data structure.

--
// SASADA Koichi at atdot dot net

[ko1 (Koichi Sasada)]

• Category set to core
• Status changed from Open to Assigned
• Assignee set to ko1 (Koichi Sasada)
• Priority changed from Normal to 3

[nahi (Hiroshi Nakamura)]

• Target version set to 2.0.0

[ko1 (Koichi Sasada)]

• Target version changed from 2.0.0 to 2.6

[ko1 (Koichi Sasada)]

• Description updated (diff)
• Status changed from Assigned to Closed

Now, we don't see this issue.