Re: V3 testsuite results (with required minor patch), FreeBSD 3.4

[Gabriel Dos Reis <gdr at codesourcery dot com>]

Loren James Rittle <rittle@latour.rsch.comm.mot.com> writes:

[...]

| I have been (mistakingly?) trying to configure and build with
| --enable-cshadow-headers (if you have completely bootstrapped with

This is still experimental and not functioning.

[...]

| Is there any good reason why I should work towards making
| --enable-cstdio=libio work as well?  It appears not.

Well, the I/O part of libstdc++ is based on libio.  Whether that is
well/ill-advised is a controversial issue and I don't have the time to
open that can of worms. On non-libio systems (Solaris for example), we
build a light version of libio necessary and sufficient for V3.  That
is you don't need to have libio on your system in order to have a
fonctioning C++ runtime system.

| If not, would an additional patch to make --enable-cstdio=stdio the
| default on FreeBSD be accepted?

Yes, if you find the default unacceptable.

| Here is the small patch required (notice that no patch was yet
| required to libstdc++-v3/config/os/bsd/bits/os_defines.h) [Would a V3
| guru please review (I think it is obvious and it has passed bootstrap
| on i686-unknown-freebsd3.4) and install for me since I have no write
| access]:
| 
| 2000-11-13  Loren J. Rittle  <ljrittle@acm.org>
| 	* config/c_io_stdio.h: Don't define __c_wfile_type
|         when _GLIBCPP_USE_WCHAR_T is not defined.

This is OK.

| Here are the testsuite results (I have not yet determined why results
| for shared are much worse than for static - it doesn't appear to be
| directly related to the third point learned the hard way by Mark
| Mitchell in http://sources.redhat.com/ml/libstdc++/2000-11/msg00139.html
| but I haven't confirmed that yet):

I need to take a closer look.  For the time being I'll install your
patch.  Could you try to run "by hand" some if the failing tests using
shared libraries?

Thanks,

-- Gaby
CodeSourcery, LLC                       http://www.codesourcery.com