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

[Andrew Haley <aph at redhat dot com>]

Rainer Orth writes:
 > Andrew Haley writes:
 > 
 > > Perhaps, although not using capitalized names is just a matter of
 > > convention.  The system is treading on the user's name space!
 > 
 > True enough.  I'll check in more detail what's going on here.  There may be
 > another problem: between the last two builds, Solaris 10 changed from the
 > older __STDC__ - 0 == 0 or similar constructs to a new macro _STRICT_STDC.
 > The problem may be due to that change.
 > 
 > > Anyway, I don't know why that file includes java-signal.h at all: it
 > > should not.
 > > 
 > > Try removing
 > > 
 > > #include <java-signal.h>
 > > 
 > > from interpret.cc.
 > 
 > ... which won't help for verify.cc.

which shouldn't include it either.

 > Besides, <sys/signal.h> is dragged in indirectly via
 > <sys/select.h>, so this won't help.

Right.  Well, this is defintely a Solaris bug.  What you seem to be
saying is that on Solaris any program that uses any stdio headers can
not define a class member or a variable called PC.

Well, okay: please put the undef at the top of the file, with an
appropriate comment.

Andrew.