On Wed, Dec 12, 2012 at 8:41 PM, Jakub Jelinek <ja...@redhat.com> wrote:
> Hi!
>
> On the attached testcase prg_ctr_mask is non-zero, presumably set
> while there still were some functions in the TU, but later on none of them
> are being emitted. This leads to n_functions in coverage_obj_finish being
> 0, and the array thus containing 0x100000000 elements.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?
Ok.
Thanks,
Richard.
> 2012-12-12 Jakub Jelinek <ja...@redhat.com>
>
> PR gcov-profile/55650
> * coverage.c (coverage_obj_init): Return false if no functions
> are being emitted.
>
> * g++.dg/other/pr55650.C: New test.
> * g++.dg/other/pr55650.cc: New file.
>
> --- gcc/coverage.c.jj 2012-11-19 14:41:24.000000000 +0100
> +++ gcc/coverage.c 2012-12-12 08:54:35.005180211 +0100
> @@ -999,6 +999,9 @@ coverage_obj_init (void)
> /* The function is not being emitted, remove from list. */
> *fn_prev = fn->next;
>
> + if (functions_head == NULL)
> + return false;
> +
> for (ix = 0; ix != GCOV_COUNTERS; ix++)
> if ((1u << ix) & prg_ctr_mask)
> n_counters++;
> --- gcc/testsuite/g++.dg/other/pr55650.C.jj 2012-12-12 09:03:53.342876593
> +0100
> +++ gcc/testsuite/g++.dg/other/pr55650.C 2012-12-12 09:03:11.000000000
> +0100
> @@ -0,0 +1,21 @@
> +// PR gcov-profile/55650
> +// { dg-do link }
> +// { dg-options "-O2 -fprofile-generate" }
> +// { dg-additional-sources "pr55650.cc" }
> +
> +struct A
> +{
> + virtual void foo ();
> +};
> +
> +struct B : public A
> +{
> + B ();
> + void foo () {}
> +};
> +
> +inline A *
> +bar ()
> +{
> + return new B;
> +}
> --- gcc/testsuite/g++.dg/other/pr55650.cc.jj 2012-12-12 09:03:56.329858741
> +0100
> +++ gcc/testsuite/g++.dg/other/pr55650.cc 2012-12-12 09:03:48.982900718
> +0100
> @@ -0,0 +1,4 @@
> +int
> +main ()
> +{
> +}
>
> Jakub
