[parisc-linux] gcc (g++) problem

[parisc-linux] gcc (g++) problem

[Jeremy Drake]

As you may remember from my last question, I'm having some problems
building some code on parisc-linux.  I have been able to reproduce the
problem using a more simple program.

Here's the problem.  This program segfaults when compiled with -fPIC, as
required to build a .so.  It is fine when compiled without -fPIC.  It is
also fine on intel platform with both.  Is there something wrong with the
code that I'm not seeing, or is this a compiler bug?  I have tried this
with gcc-3 on an intel box, and it is fine, so I assume it is this port?

Thanks for your help...

g++ -g -o test -fPIC test.cpp

---- test.cpp ----
#include <stdio.h>

class Foo
{
public:
	Foo(const char * const blah);
	unsigned char* text;
};
Foo::Foo(const char * const blah)
{
	text=(unsigned char *)blah;
}

const Foo test("Blah");
int main()
{
	printf("%s\n", test.text);
	return 0;
}



-- 
Advertisements contain the only truths to be relied on in a newspaper.
		-- Thomas Jefferson

Re: [parisc-linux] gcc (g++) problem

[John David Anglin]

> Here's the problem.  This program segfaults when compiled with -fPIC, as
> required to build a .so.  It is fine when compiled without -fPIC.  It is
> also fine on intel platform with both.  Is there something wrong with the
> code that I'm not seeing, or is this a compiler bug?  I have tried this
> with gcc-3 on an intel box, and it is fine, so I assume it is this port?

It's probably a compiler bug.  The code works ok with my 3.1 test version.
I think the problem may be the restoration of the pic offset table register
after function calls.

Suggest trying the debian 3.1 prerelease build.  Search the list to find
how to get it.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: [parisc-linux] gcc (g++) problem

[Randolph Chung]

> It's probably a compiler bug.  The code works ok with my 3.1 test version.
> I think the problem may be the restoration of the pic offset table register
> after function calls.
> 
> Suggest trying the debian 3.1 prerelease build.  Search the list to find
> how to get it.

something is broken in the debian 3.1 snapshot package,
unfortunately....

legolas[20:35] ~% /usr/lib/gcc-snapshot/bin/g++ -o foo foo.cc 
/usr/bin/ld: unrecognized option '--eh-frame-hdr'
/usr/bin/ld: use the --help option for usage information
collect2: ld returned 1 exit status

randolph
-- 
   @..@                                         http://www.TauSq.org/
  (----)
 ( >__< )
 ^^ ~~ ^^

Re: [parisc-linux] gcc (g++) problem

[John David Anglin]

> > Suggest trying the debian 3.1 prerelease build.  Search the list to find
> > how to get it.
> 
> something is broken in the debian 3.1 snapshot package,
> unfortunately....

No, the problem is 3.1 was built with a recent version of binutils
which supports --eh-frame-hdr.  This change also involved changing
the way libgcc and the shared version of libgcc were linked with
shared libraries.  Do you recall your report of a build failure of
3.1 linking libstdc++?  My trying to fix this problem lead to a
heated discussion on gcc-patches.  My fix was reverted, then further
problems arose, and a new patch affecting the linkage of libgcc
was installed yesterday.

> legolas[20:35] ~% /usr/lib/gcc-snapshot/bin/g++ -o foo foo.cc 
> /usr/bin/ld: unrecognized option '--eh-frame-hdr'
> /usr/bin/ld: use the --help option for usage information
> collect2: ld returned 1 exit status

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)

Re: [parisc-linux] gcc (g++) problem

[Jeremy Drake]

Cool.  Seems to work OK on the test, now to try it on the real thing...


On Mon, 25 Feb 2002, John David Anglin wrote:

> > Here's the problem.  This program segfaults when compiled with -fPIC, as
> > required to build a .so.  It is fine when compiled without -fPIC.  It is
> > also fine on intel platform with both.  Is there something wrong with the
> > code that I'm not seeing, or is this a compiler bug?  I have tried this
> > with gcc-3 on an intel box, and it is fine, so I assume it is this port?
>
> It's probably a compiler bug.  The code works ok with my 3.1 test version.
> I think the problem may be the restoration of the pic offset table register
> after function calls.
>
> Suggest trying the debian 3.1 prerelease build.  Search the list to find
> how to get it.
>
> Dave
>

-- 
QOTD:
	 "It's not the despair... I can stand the despair.  It's the hope."