This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: HP's compiler crashes when building gcc from trunk


On Wed, Sep 03, 2003 at 05:30:45PM -0700, I wrote:
> I just tried building the trunk from an HP/UX system, using the HP-supplied
> C compiler.  This has worked fine for 3.2.x and 3.3.x, but the trunk
> crashes HP's compiler:
> 
> cc -c   -g  -DIN_GCC     -DHAVE_CONFIG_H    -I. -I. -I/u/jbuck/gnu/src/cvs/trunk/gcc/gcc -I/u/jbuck/gnu/src/cvs/trunk/gcc/gcc/. -I/u/jbuck/gnu/src/cvs/trunk/gcc/gcc/../include -I../intl /u/jbuck/gnu/src/cvs/trunk/gcc/gcc/c-decl.c -o c-decl.o
> cc: line 508: panic 5172: Backend Assert ** Unimplemented CVT. (5172)
> 
> The versions I have are
> 
> cpp.ansi: HP92453-01 B.11.11.04 HP C Preprocessor (ANSI)
> ccom: HP92453-01 B.11.11.04 HP C Compiler
> 
> Anyone seen this before?

Following up my own report: the bug in HP's compiler has to do with the
line

	scope->function_body |= functionbody;

where function_body is a 1-bit bitfield and functionbody is an int.
Evidently HP didn't bother to implement support for this operation.
If I change the line to

	scope->function_body = scope->function_body | functionbody;

the file compiles.  I haven't gotten through a full bootstrap yet.

It seems that having the compiler bootstrap is important enough to
justify this ugliness.  What do folks think?  Put this in as a patch?
I'll wait to verify that I can bootstrap and test the compiler, of
course.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]