On 5/23/24 09:27, Nathaniel Shead wrote:
On Mon, May 20, 2024 at 06:00:09PM -0400, Jason Merrill wrote:
On 5/17/24 02:14, Nathaniel Shead wrote:
On Tue, May 14, 2024 at 06:21:48PM -0400, Jason Merrill wrote:
On 5/12/24 22:58, Nathaniel Shead wrote:
Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for trunk?
OK.
I realised as I was looking over this again that I might have spoken too
soon with the header unit example being supported. Doing the following:
// a.H
struct { int y; } s;
decltype(s) f(decltype(s)); // { dg-error "used but never defined" }
inline auto x = f({ 123 });
// b.C
struct {} unrelated;
import "a.H";
decltype(s) f(decltype(s) x) {
return { 456 + x.y };
}
// c.C
import "linkage-3_a.H";
int main() { auto a = x.y; }
Actually does fail to link, because in 'c.C' we call 'f(.anon_0)' but
the definition 'b.C' is f(.anon_1).
I don't think this is fixable, so I don't think this direction is
workable.
Since namespace-scope anonymous types are TU-local, we don't need to support
that for proper modules, but it's not clear to me that we don't need to
support it for header units.
OTOH, https://eel.is/c++draft/module#import-5.3 allows c.C to import a
different header unit than b.C, in which case the type is different and x
violates the odr.
Right; I think at this stage I don't know how to support this for header
units (and even for module interface units it doesn't actually work;
more on this below), so I think saying that this is actually an ODR
violation is OK.
That said, I think that it might still be worth making header modules
satisfy 'module_has_cmi_p', since that is true to the name, and will be
useful in other places we currently use 'module_p ()': in which case we
could instead make all the callers in 'no_linkage_check' do
'module_maybe_has_cmi_p () && !header_module_p ()'; something like the
following, perhaps?
If we need that condition, it should be its own predicate rather than
expecting callers to do that combined check.
But it's not clear to me how this is different from a type in the GMF of a
named module, which is exactly the maybe_has_cmi case; there we could again
see a different version of the type if another TU includes the header.
Jason
This made me go back and double-check for named modules and it actually
does fail as well; the following sample ICEs, even:
export module M;
struct {} s;
int h(decltype(s));
int x = h(s); // ICE in write_unnamed_type_name, cp/mangle.cc:1806
So I think maybe the way to go here is to just not treat unnamed types
as something that could possibly be accessed from a different TU, like
the below. Then we don't need to do the special handling for header
units, since as you say, they're not materially different anyway.
Thoughts?
Sounds good.
(And I've moved the original change to 'module_has_cmi_p' to a separate
patch given it's somewhat unrelated now.)
Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for trunk (and
maybe 14.2)?
-- >8 --
In r14-9530 we relaxed "depending on type with no-linkage" errors for
declarations that could actually be accessed from different TUs anyway.
However, this also enabled it for unnamed types, which never work.
In a normal module interface, an unnamed type is TU-local by
[basic.link] p15.2, and so cannot be exposed or the program is
ill-formed. We don't yet implement this checking but we should assume
that we will later; currently supporting this actually causes ICEs when
attempting to create the mangled name in some situations.
For a header unit, by [module.import] p5.3 it is unspecified whether two
TUs importing a header unit providing such a declaration are importing
the same header unit. In this case, we would require name mangling
changes to somehow allow the (anonymous) type exported by such a header
unit to correspond across different TUs in the presence of other
anonymous declarations, so for this patch just assume that this case
would be an ODR violation instead.
diff --git a/gcc/testsuite/g++.dg/modules/linkage-2.C
b/gcc/testsuite/g++.dg/modules/linkage-2.C
index eb4d7b051af..f69bd7ff728 100644
--- a/gcc/testsuite/g++.dg/modules/linkage-2.C
+++ b/gcc/testsuite/g++.dg/modules/linkage-2.C
@@ -13,14 +13,15 @@ namespace {
return A{};
}
decltype(f()) g(); // { dg-error "used but never defined" }
-
- struct {} s;
- decltype(s) h(); // { dg-error "used but never defined" }
}
export void use() {
g();
- h();
The above changes seem undesirable; we should still give that error.
+// Additionally, unnamed types have no linkage but are also TU-local,
+// and thus cannot be in a module interface unit.
+// We don't yet implement this checking however.
+struct {} s; // { dg-error "TU-local" "" { xfail *-*-* } }
The comment should clarify that the problem is the non-TU-local variable
's' exposing the TU-local type. If s is also TU-local (as the one in
the anonymous namespace above) it's fine as long as s is not itself exposed.
Why doesn't the exposure code catch this? I guess the code in
make_dependency to mark internal types needs to handle anonymous types
as well.
Jason
