While fixing PR algol68/123653 I noticed that there was a struct mode
interned in the compiler with the form `struct (ref int)'. This is
odd because fields are supposed to have names in struct modes.
Turns out that the culprit is:
(void) a68_add_mode_to_pack (&z, M_REF_INT, NO_TEXT, NO_NODE);
EQUIVALENT (M_SEMA) = a68_add_mode (&TOP_MOID (&A68_JOB),
STRUCT_SYMBOL,
a68_count_pack_members (z),
NO_NODE, NO_MOID, z);
i.e. it is the standard mode `sema'. The report says that it is a
struct that hold a reference to int within, but you are not allowed to
access it.
This patch adds a comment to are_packs_equivalent explaining why the
name of a pack may be NULL.
Signed-off-by: Jose E. Marchesi <[email protected]>
gcc/algol68/ChangeLog
* a68-parser-moids-equivalence.cc (are_packs_equivalent): Add
explicative comment.
---
gcc/algol68/a68-parser-moids-equivalence.cc | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/gcc/algol68/a68-parser-moids-equivalence.cc
b/gcc/algol68/a68-parser-moids-equivalence.cc
index 260eb4209a6..f97dcafc694 100644
--- a/gcc/algol68/a68-parser-moids-equivalence.cc
+++ b/gcc/algol68/a68-parser-moids-equivalence.cc
@@ -49,6 +49,10 @@ are_packs_equivalent (PACK_T *s, PACK_T *t,
return false;
if (compare_names)
{
+ /* Note that a name in a pack of a struct may be NULL if the
+ corresponding field is supposed to not be accessible by the user.
+ There is at present one instance of this: the standard prelude
+ */
if (TEXT (s) != TEXT (t)
&& !(TEXT (s) != NO_TEXT
&& TEXT (t) != NO_TEXT
--
2.39.5
