Re: 3.4/3.5 PATCH: Fix Solaris 10 libjava bootstrap failure

[Rainer Orth <ro at TechFak dot Uni-Bielefeld dot DE>]

Andrew Haley writes:

> Rainer Orth writes:
>  > 
>  > Only signal.h.
> 
> But who includes that?  In Linux it seems to be included by pthread.h,
> which comes via java-thread.h.  It seems likely that you'll end up
> including signal.h by accident.

-H reveals:

[...]
.. ./include/java-threads.h
... /usr/include/pthread.h
.... /usr/include/sys/feature_tests.h
..... /usr/include/sys/isa_defs.h
.... /vol/gcc/obj/gcc-3.4.0-20040318/10-gcc/gcc/include/sys/types.h
..... /usr/include/sys/machtypes.h
...... /usr/include/ia32/sys/machtypes.h
..... /usr/include/sys/int_types.h
..... /usr/include/sys/select.h
...... /vol/gcc/obj/gcc-3.4.0-20040318/10-gcc/gcc/include/sys/signal.h
[...]

I.e. <pthread.h> includes <sys/types.h>, which again includes
<sys/select.h> with this comment:

#if !defined(__XOPEN_OR_POSIX) || defined(__EXTENSIONS__)
[...]
/*
 * Nested include for BSD/sockets source compatibility.
 * (The select macros used to be defined here).
 */
#include <sys/select.h>

The latter includes <sys/signal.h> like so:

#if !defined(__XOPEN_OR_POSIX) || defined(_XPG6) || defined(__EXTENSIONS__)
#include <sys/signal.h>
#include <sys/time_impl.h>
#endif

As you can see, defining __EXTENSIONS__ is the source of all evil.  It
might be useful to try and get rid of this, but I fear this opens several
cans of worms ;-(

	Rainer

-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University