Re: [PATCH] Avoid polluting cygwin namespace.

[Bruno Haible]

Eric Blake wrote:
> >    error: redefinition of typedef ‘CONTEXT’
>
> That can be worked around; here's an alternative approach:
> 
> FAULT_CONTEXT='struct _CONTEXT'
> FAULT_CONTEXT_INCLUDE='struct _CONTEXT; /* include <windows.h> to see inside 
> */'

Yes, this would fix the first problem.

> > 2) The users of libsigsegv need to access the field of the FAULT_CONTEXT.
> >    That's what it is provided for, in the first place. For example, GNU 
> > clisp
> >    does in src/spvw_sigsegv.d:
> > 
> >      stackoverflow_context_t scp = (stackoverflow_context_t) arg1;
> >      ...
> >      #if defined(WIN32_NATIVE) || defined(UNIX_CYGWIN32)
> >       #ifdef I80386
> >        if (scp) { setSTACK(STACK = (gcv_object_t*)(scp->Ebx)); }
> >       #endif
> >      #endif
> > 
> >    This would be impossible if stackoverflow_context_t was an opaque type.
> 
> If you are already using windows-specific #ifdefs, then it doesn't hurt
> to add another one in src/spvw_sigsegv.d:
> 
> #if defined WIN32_NATIVE || defined UNIX_CYGWIN32
> # include <windows.h>
> #endif

It does hurt. libsigsegv is supposed to take away a maximum of portability
problems from the user. On some platforms, you need <signal.h> to get
the type defined, or <asm/sigcontext.h>, on some others <ucontext.h>, or
maybe some <mach/*> header, etc.

And when Cygwin implements <ucontext.h>, libsisegv users should not have
to change their #include <windows.h> to #include <ucontext.h>. libsigsegv
is designed to take away that worry.

> for cygwin, it goes
> counter to the goals of cygwin of providing a Unix-y environment.  True,
> cygwin does not yet provide ucontext_t

So that is the problem, and that needs to be fixed. Cygwin requires
including <windows.h> for some services that are in the scope of POSIX.
Why would you propose to change libsigsegv because of this? It would
not solve the real problem.

Bruno