Re: MPS codegen

["Gerd Möllmann" <gerd.moellmann@gmail.com>]

Gerd Möllmann <gerd.moellmann@gmail.com> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> I still think we should first establish that that is indeed the problem.
>>> Please see my other mail.
>>
>> Since you insist: I ifdef'ed away this:
>>
>>   #if 0
>> 	/* Release the commit charge for the mapping.  */
>> 	DWORD old_prot;
>> 	(void) VirtualProtect (mem, size, PAGE_NOACCESS, &old_prot);
>>   #endif
>>
>> and, lo and behold, Emacs can now start without a problem.
>>
>> I hope this establishes the fact that we are still accessing memory
>> from the "discarded" pdumper block.
>>
>> Now what?
>
> Thanks, and sorry, but that felt too much like guessing to me :-). I'll
> see if I can come up with something equivalent on my system.

Erm, I can't - 

  void
  dump_discard_mem (void *mem, size_t size)
  {
  #if VM_SUPPORTED == VM_MS_WINDOWS
        /* Discard COWed pages.  */
        (void) VirtualFree (mem, size, MEM_DECOMMIT);
        /* Release the commit charge for the mapping.  */
        DWORD old_prot;
        (void) VirtualProtect (mem, size, PAGE_NOACCESS, &old_prot);
  #elif VM_SUPPORTED == VM_POSIX
  # ifdef HAVE_POSIX_MADVISE
        /* Discard COWed pages.  */
        (void) posix_madvise (mem, size, POSIX_MADV_DONTNEED);
  # elif defined HAVE_MADVISE
        (void) madvise (mem, size, MADV_DONTNEED);
  #endif
        /* Release the commit charge for the mapping.  */
        (void) mprotect (mem, size, PROT_NONE);
  #endif
  }

This does already an mprotect PROT_NONE on my system (no permissons at
all), and there is nothing nore I can do. I can't munmap because that
would write back the changes done by the reloc phase back to the .pdmp
file.