Re: MPS: unable to build due to assertion violation in igc_dump_check_object_starts

[Pip Cet <pipcet@protonmail.com>]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

On Wednesday, July 24th, 2024 at 09:39, Helmut Eller <eller.helmut@gmail.com> wrote:
> On Wed, Jul 24 2024, Pip Cet wrote:
> 
> > Is alignof(struct Lisp_Vector) == 4 for you?
> 
> No, it's 8. Because of the GCALIGNED_STRUCT attribute.

Thanks for letting me know.  Is this in gdb or lldb?  lldb shows me alignof(struct Lisp_Vector) == 4 on a similar Debian system, which looks wrong to me.  I've fixed things on that system for now (it was building without mps, because that requires -pthread in the CFLAGS to link on that machine...)

Can you try the attached patch?

Pip

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: 0001-mps-aligned.patch --]
[-- Type: text/x-patch; name=0001-mps-aligned.patch, Size: 904 bytes --]

diff --git a/src/lisp.h b/src/lisp.h
index 8a1b1acf5f0..d4331b3a4ff 100644
--- a/src/lisp.h
+++ b/src/lisp.h
@@ -45,7 +45,7 @@ #define EMACS_LISP_H
 INLINE_HEADER_BEGIN
 
 #ifdef HAVE_MPS
-union gc_header { uint64_t v; };
+union gc_header;
 extern void gc_maybe_quit (void);
 extern bool gc_signal_handler_can_run (int);
 #else
@@ -335,6 +335,7 @@ #define GCALIGNED_UNION_MEMBER char alignas (GCALIGNMENT) gcaligned;
 #endif
 #define GCALIGNED(type) (alignof (type) % GCALIGNMENT == 0)
 
+union gc_header { uint64_t v; GCALIGNED_UNION_MEMBER };
 /* Lisp_Word is a scalar word suitable for holding a tagged pointer or
    integer.  Usually it is a pointer to a deliberately-incomplete type
    'struct Lisp_X'.  However, it is EMACS_INT when Lisp_Objects and
@@ -6335,3 +6336,4 @@ define_error (Lisp_Object name, const char *message, Lisp_Object parent);
 INLINE_HEADER_END
 
 #endif /* EMACS_LISP_H */
+