lloda pushed a commit to branch main
in repository guile.
commit 9b00131539ca11a8112a4e88187e45bd40a6f7a7
Author: Rob Browning <[email protected]>
AuthorDate: Tue Sep 9 12:15:54 2025 -0500
make-struct/no-tail scm_c_make_structv: ensure layout is symbol
Previously we just tested for a layout via scm_is_true (SCM_LAYOUT ...),
but layouts are symbols, and that test allowed us to proceed to the code
in scm_is_valid_vtable_layout that called scm_i_symbol_chars on the
"layout", segfaulting if the layout was say an integer, as it would be
for something like this:
(make-struct/no-tail (make-vtable standard-vtable-fields #f) 0)
Since a layout must be a symbol, check via scm_is_symbol instead.
* libguile/struct.c (scm_c_make_structv): test for layout via
scm_is_symbol.
Thanks to Ido Yariv for reporting the problem and Dale P. Smith for help
determining the cause.
Closes: 78789
---
NEWS | 4 ++++
libguile/struct.c | 2 +-
2 files changed, 5 insertions(+), 1 deletion(-)
diff --git a/NEWS b/NEWS
index 264c16322..d57c342a0 100644
--- a/NEWS
+++ b/NEWS
@@ -205,6 +205,10 @@ change.
** A numer of test incompatibilities with Darwin hosts have been fixed
(<https://bugs.gnu.org/72547>)
** define guile-test as non-declarative because it uses load
+** make-struct/no-tail and scm_c_make_structv no longer segfault
+ Previously they could if given the standard-vtable-fields, but inital
+ values that weren't compatible with a vtable.
+ (<https://bugs.gnu.org/78789>)
Changes in 3.0.10 (since 3.0.9)
diff --git a/libguile/struct.c b/libguile/struct.c
index 68dcc0070..ef2225e6f 100644
--- a/libguile/struct.c
+++ b/libguile/struct.c
@@ -362,7 +362,7 @@ scm_c_make_structv (SCM vtable, size_t n_tail, size_t
n_init, scm_t_bits *init)
initialization, to humor GOOPS, so only validate if the layout was
passed as an initarg. */
if (SCM_VTABLE_FLAG_IS_SET (vtable, SCM_VTABLE_FLAG_VTABLE)
- && scm_is_true (SCM_VTABLE_LAYOUT (obj)))
+ && scm_is_symbol (SCM_VTABLE_LAYOUT (obj)))
scm_i_struct_inherit_vtable_magic (vtable, obj);
return obj;
