This is the mail archive of the 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]

Re: egcs 1.0.1 problems under Irix 6.3

	As you are probably already aware, egcs has a lot of problems under Irix

I have an irix 6.3 machine on my desk.  I have not seen a lot of problems.
There is only one problem that I know of, an ABI issue where gcc emits code
incompatible with SGI C in some rare cases.

However, I am not a C++ user, so there could be C++/irix6 interaction problems
that I am not aware of, but I haven't seen many bug reports, so I am
skeptical.  If you are a heavy C++ user, you might be running into generic
C++ problems that have nothing to do with irix6.

	OString& OString::operator += ( char ch )
	  assert( assigned() );
	  assert( ch != 0 );

	  realloc( theLength + 5 );	// mp: this is line 223
	  theString[theLength++]= ch;
	  theString[theLength]= 0;
	  return *this;

If you are going to give us testcases, you need to give us ones that we can
compile, plus all info we need to reproduce the problem, such as the gcc
command you typed, and what configure options you used if any.

This is not a valid testcase, because there are no declarations for any of
the types or variables used here, so I can't do anything with it.
I don't have time to try to download an entire package and try to build it.

	In addition, ld gives me segmentation faults without fail when running g++
	with the flag '-s'.  ld's man page has this to say about the '-s' option:
	     -s   Strip the symbolic information from the output object file.

Interesting.  It is trivial to reproduce this.  There is obviously a bug in
SGI's linker, as it shouldn't be core dumping.  There might also be a bug
in gcc, but this may be hard to track down without help from SGI.

	I've found both egcs and gcc 2.8.x to be essentially unusable under Irix 6.3
	with IDO 7.1.

Why?  I have the same thing on my desk, and it works for me.  Details please.

	I have had some success with egcs 1.0.1 under Irix 6.2 and
	IDO 7.0.1, but of course those systems implement the MIPS III ISA rather
	than the MIPS IV ISA.

The underlying hardware should not make any difference.    Gcc will emit the
same code by default regardless of which version of Irix6 you have, or which
version of the IDO you have, or what kind of hardware you have.

	I suspect that at issue is egcs's lax enforcement of
	its exclusively O32 ABI linking policy.

This doesn't make any sense to me.  The egcs irix6 port emits N32 code by
default, and does not support O32 code at all.  I also have no idea what
you mean by `lax enforcement'.


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