This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: GCC 3.2 branch libstdc++ abi_check.cc bootstrap failure on AIX
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Benjamin Kosnik <bkoz at redhat dot com>
- Cc: mark at codesourcery dot com, gcc-bugs at gcc dot gnu dot org
- Date: Thu, 24 Oct 2002 13:39:59 -0400
- Subject: Re: GCC 3.2 branch libstdc++ abi_check.cc bootstrap failure on AIX
>>>>> Benjamin Kosnik writes:
Ben> so... if you revert this, does the ICE go away? I would be surprised if it did.
Ben> This is pretty simple C++ code, and under 400 lines. If this cannot be
Ben> built (and linked), something is seriously wrong.
Yes, if I revert just the abi_check.cc changes, abi_check.cc
compiles without the reload ICE.
The ICE only occurs for -mcpu=power, but does not occur on the
trunk with that same option. I am not sure if anyone has run a regression
check with -mcpu=power recently, but the rest of libstdc++.a does compile
with -mcpu=power.
-mcpu=power is no longer supported with AIX 5.2 and above. I do
not know if anyone uses that option with G++ -- I have no received any bug
reports.
The issue is that -mcpu=power is not an important sub-target worth
spending a lot of time tracking down the cause of the reload ICE.
However, this bootstrap failure did not occur before the abi_check.cc
change and does not occur on the trunk. Disabling support for -mcpu=power
in an update release to hide this problem under the rug is not
appropriate.
With the abi_check.cc change, users can no longer download the GCC
3.2.1 release, configure it for AIX 4.3 or AIX 5.1, and have it
bootstrap. If this were just a new regression, I wouldn't care, but not
being able to bootstrap the release out of the box is a problem.
David