Re: Possible patch for fortran/68078

[Fritz Reese <fritzoreese@gmail.com>]

It might be worth filling in the right locations for the new blocks (?):

diff --git a/gcc/fortran/resolve.c b/gcc/fortran/resolve.c
@@ -7048,7 +7048,9 @@ cond_init (gfc_code *code, gfc_expr *e, int pointer, gfc_e
   init_st->expr2 = init_e;

   block = gfc_get_code (EXEC_IF);
+  block->loc = code->loc;
   block->block = gfc_get_code (EXEC_IF);
+  block->block->loc = code->loc;
   block->block->expr1 = cond;
   block->block->next = init_st;
   block->next = code->next;


Minor nit, trailing whitespace:

--- gcc/fortran/resolve.c       (revision 240152)
+++ gcc/fortran/resolve.c       (working copy)
...
+  block->block->expr1 = cond;
+  block->block->next = init_st;
+  block->next = code->next;
+  <<<<< Trailing whitespace.
+  code->next = block;

Otherwise it looks fine to me. So long as the system(s) to test
support(s) setrlimit() (which should be true for POSIX systems).

---
Fritz Reese

(P.S. I can't officially approve the patch, I'm just providing feedback.)

On Thu, Sep 15, 2016 at 2:11 AM, Louis Krupp <louis.krupp@zoho.com> wrote:
> Verify that allocation was successful before assigning default initialization values to components of the allocated object (or array of > objects or pointer to an object).   The change could probably have been made while parsing or at translation.  This patch does it in > the resolution phase.  It seemed like a good idea.
>