On Tue, Sep 01, 2020 at 03:27:36PM -0400, Jason Merrill via Gcc-patches wrote:
> On 9/1/20 12:10 PM, Marek Polacek wrote:
> > Currently, we allow new char[]{"foo"}, but not new char[4]{"foo"}.
> > We should accept the latter too: [dcl.init.list]p3.3 says to treat
> > this as [dcl.init.string].
> >
> > We were rejecting this code because we never called reshape_init before
> > the digest_init in build_new_1. reshape_init handles [dcl.init.string]
> > by unwrapping the STRING_CST from its enclosing { }, and digest_init
> > assumes that reshape_init has been called for aggregates anyway, and an
> > array is an aggregate.
> >
> > Bootstrapped/regtested on x86_64-pc-linux-gnu, ok for trunk?
> >
> > gcc/cp/ChangeLog:
> >
> > PR c++/77841
> > * init.c (build_new_1): Call reshape_init.
> >
> > gcc/testsuite/ChangeLog:
> >
> > PR c++/77841
> > * g++.dg/cpp0x/initlist-new4.C: New test.
> > ---
> > gcc/cp/init.c | 6 ++++++
> > gcc/testsuite/g++.dg/cpp0x/initlist-new4.C | 6 ++++++
> > 2 files changed, 12 insertions(+)
> > create mode 100644 gcc/testsuite/g++.dg/cpp0x/initlist-new4.C
> >
> > diff --git a/gcc/cp/init.c b/gcc/cp/init.c
> > index 360ab8c0b52..d4540db3605 100644
> > --- a/gcc/cp/init.c
> > +++ b/gcc/cp/init.c
> > @@ -3575,6 +3575,12 @@ build_new_1 (vec<tree, va_gc> **placement, tree
> > type, tree nelts,
> > /* We'll check the length at runtime. */
> > domain = NULL_TREE;
> > arraytype = build_cplus_array_type (type, domain);
> > + /* If we have new char[4]{"foo"}, we have to reshape
> > + so that the STRING_CST isn't wrapped in { }. */
> > + vecinit = reshape_init (arraytype, vecinit, complain);
> > + /* The middle end doesn't cope with the location wrapper
> > + around a STRING_CST. */
> > + STRIP_ANY_LOCATION_WRAPPER (vecinit);
> > vecinit = digest_init (arraytype, vecinit, complain);
> > }
>
> This is OK, but now I wonder why your earlier addition,
>
> > /* This handles code like new char[]{"foo"}. */
> > else if (len == 1
> > && char_type_p (TYPE_MAIN_VARIANT (type))
> > && TREE_CODE (tree_strip_any_location_wrapper
> > ((**init)[0]))
> > == STRING_CST)
> > {
> > vecinit = (**init)[0];
> > STRIP_ANY_LOCATION_WRAPPER (vecinit);
> > }
>
> isn't handled by the block you're changing in this patch. Why isn't
> DIRECT_LIST_INIT_P true for new char[]{"foo"}?
Yes, I was hoping this hunk would handle the new char[4]{"foo"} case too,
but for new char[]{"foo"} DIRECT_LIST_INIT_P is false because earlier in
build_new we called reshape_init:
4011 /* Otherwise we should have 'new T[]{e_0, ..., e_k}'. */
4012 if (BRACE_ENCLOSED_INITIALIZER_P (elt))
4013 elt = reshape_init (type, elt, complain);
which unwraps the { } from {"foo"}, so it's no longer a list init. We
won't get there with new char[4]{"foo"} because TREE_CODE (type) will
not be ARRAY_TYPE; instead, nelts is set to INTEGER_CST 4 when we know
the array bound.
I could make it so that we call reshape_init in build_new for the [4]
case too, but it was uglier than this fix.
Should I go ahead with this patch as-is or would you prefer any changes?
Marek
