[Bug other/110885] New: "undefined reference" when using both "-fdebug-types-section" and "-gsplit-dwarf"
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110885 Bug ID: 110885 Summary: "undefined reference" when using both "-fdebug-types-section" and "-gsplit-dwarf" Product: gcc Version: unknown Status: UNCONFIRMED Severity: normal Priority: P3 Component: other Assignee: unassigned at gcc dot gnu.org Reporter: linh.lrd at gmail dot com Target Milestone: ---
[Bug debug/110885] "undefined reference" when using both "-fdebug-types-section" and "-gsplit-dwarf"
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110885
--- Comment #1 from Ling Nguyen ---
Consider this program:
struct X
{
static void foo() {}
};
struct Y
{
template
static void bar() {}
};
int main()
{
Y::bar();
return 0;
}
If it's compiled using something like: gcc -gsplit-dwarf -fdebug-types-section.
the linker will complain: undefined reference to `X::foo()'
More details here: https://godbolt.org/z/qnj56e769
[Bug debug/110885] [10/11/12/13/14 Regression] "undefined reference" when using both "-fdebug-types-section" and "-gsplit-dwarf" and inline function as a template argument
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110885
--- Comment #3 from Ling Nguyen ---
Thanks for quick responses.
A further look into the issue:
I believe there is bug in this piece of code:
https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/dwarf2out.cc;h=fa0fe4c41bbd5b13aa5fe6cad6e7d39bbd82174a;hb=HEAD#l8501
It should probably be:
if (is_template_instantiation (c))
{
/* Instantiated templates do not need to be cloned into the
type unit. Just move the DIE and its children back to
the skeleton tree (in the main CU). */
remove_child_with_prev (c, prev);
generate_skeleton_ancestor_tree (parent);
add_child_die (parent->new_die, c);
c = prev;
}
Don't mean to teach a fish to swim here, because I'm no expert. Just want to
help to reduce the search space.
By the way, any chance for a fix (whenever it is) to land on gcc 10.2, as we
are fixed to it?
