[Mono-dev] /mono/mini/main.c build error: depends on HAVE_SGEN_GC define, making it impossible to compile for sgen

[Jeremy Bell]

At some point between branch mono-2-10-9 and branch master, a change was
made to /mono/mini/main.c:

branch mono-2-10-9:

main.c:
#include <config.h>
#include "mini.h"
#ifndef HOST_WIN32
#include "buildver.h"
#endif


branch master:
#include <config.h>
#include "mini.h"
#ifndef HOST_WIN32
#ifdef HAVE_SGEN_GC
#include "buildver-sgen.h"
#else
#include "buildver.h"
#endif
#endif

This makes main.c impossible to compile when buildver-sgen.h is generated
and not buildver.h. First of all, HAVE_SGEN_GC is never defined for files
in /mini as far as I can tell, so main.c always attempts to include
buildver.h, which does not exist when buildver-sgen.h is generated instead.

However, even if you explicitly define HAVE_SGEN_GC in CFLAGS, etc... then
you will still get an error, in mini.h, because it believes it is an error
to have either HAVE_SGEN_GC or HAVE_BOEHM_GC defined when mini.h is
included, as /mini code should not have dependencies on the GC being used,
so it says:

mini.h:
/*
 * The mini code should not have any compile time dependencies on the GC
being used, so the same object file from mini/
 * can be linked into both mono and mono-sgen.
 */
#if defined(HAVE_BOEHM_GC) || defined(HAVE_SGEN_GC)
#error "The code in mini/ should not depend on these defines."
#endif


So, either way, main.c won't compile without modification. Is the error in
/mono/mini/mini.h no longer valid? Or was the change to /mono/mini/main.c
to depend on the HAVE_SGEN_GC define a regression?

Thanks,
Jeremy
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ximian.com/pipermail/mono-devel-list/attachments/20130523/847b2fe6/attachment.html>