Re: nonreturning abort() in MS-VC + SP_MALLOC changes break compi lation

[Hoehle, Joerg-C. <Joe...@t-...>]

Hi,

>How can this be possible? SP_bound is defined ifndef NO_SP_CHECK, and
>is used only if NO_SP_CHECK.
>
>> spvw.d was compiled as follows:
>>         ..\utils\gcc-cccp\cccp -U__GNUC__ -+ -D_M_IX86=3D500 =
-D_WIN32
>> -I"C:/Programme/Microsoft Visual Studio/VC98"/include =
-D_MSC_VER=3D1100
>> -D_INTEGRAL_MAX_BITS=3D64  -DUNICODE -DDYNAMIC_FFI=20
>-DNO_GETTEXT -DNO_SIGSEGV
>> -I. spvw.c > spvw.i.c
>
>NO_SIGSEGV implies !defined(HAVE_STACK_OVERFLOW_RECOVERY) which =
implies
>NO_SP_CHECK. Can you look (in spvw.i.c) where the unexpected reference
>to SP_bound comes from?

spvw.d:2564ff
          #if defined(WIN32_NATIVE) && !defined(NO_SP_CHECK)
            # Even if the NOCOST_SP_CHECK stack overflow detection =
(using a
            # guard page) works, we set SP_bound.
The comment looks like it should have been !defined(NOCOST_SP*)

But this is not the branch that's taken.

Oh, I see, you forgot to commit spvw.d.

BTW, lispbibl.d has the following:
#if defined(HAVE_STACK_OVERFLOW_RECOVERY)
  #define NOCOST_SP_CHECK
#else
  #define NO_SP_CHECK
#endif

#if !(defined(NO_SP_CHECK) || defined(NOCOST_SP_CHECK))
This branch is now never taken.

BTW, it seems to me like a bad idea to have to rely on the OS (or =
MS-Windows) or on the SDK editbin program to get an adequate stack =
setting, instead of a per-application user-settable variable on startup =
of CLISP.
On UNIX, one can use ulimit in the shell that starts CLISP, but what on =
MS-Windows?

What will users running stack intensive applications do?

Regards,
	J=C3=B6rg H=C3=B6hle.

Regards,
	J=C3=B6rg H=C3=B6hle