Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 08 Apr 2004 13:49:09 -0600
From:      "P.D. Seniura" <pdseniura@techie.com>
To:        freebsd-current@freebsd.org, freebsd-ports@freebsd.org
Subject:   more serious problems using lang/gcc33 or lang/gcc34 for world & kernel & some ports
Message-ID:  <20040408194909.8AAD1790058@ws1-14.us4.outblaze.com>

next in thread | raw e-mail | index | archive | help

To add to the problems mentioned previously in this thread <http://docs.FreeBSD.org/cgi/mid.cgi?20040402210539.70E945C3B>.

For the below, I'm using solely lang/gcc33 (aka gcc334).

The buildworld with NO_DYNAMICROOT=yes set in /etc/make.conf did help a bit with the previously-mentioned problems.  At least the shells found their symbols, so the system comes up to a ttyv fine.  Starting a plain KDE/X11 environment works, too.

Compiling world with NO_DYNAMICROOT=yes did _not_ help with the "undeclared MD5_*" problems when it came across libopie and others as mentioned previously.

Yesterday I removed NO_DYNAMICROOT=yes so we could again see how gcc334 does things that way.  I also applied CTM buckets thru late yesterday evening (CDT).  Now it breaks building libstdc++ as shown here:
[...]
===> gnu/lib/libstdc++
/usr/local/bin/g++33  -march=pentium2 -pipe -O2  -march=pentium2 -pipe -O2 -fno-implicit-templates -ffunction-sections -fdata-sections -Wno-deprecated -c /src/contrib/libstdc++/src/bitset.cc
[...]
/usr/local/bin/gcc33 -fpic -DPIC  -march=pentium2 -pipe -O2  -march=pentium2 -pipe -O2 -march=pentium2 -DIN_GLIBCPP_V3 -DHAVE_CONFIG_H -I/src/gnu/lib/libstdc++ -I/src/gnu/lib/libstdc++/../../../contrib/libstdc++/libsupc++ -I/src/gnu/lib/libstdc++/../../../contrib/gcc  -c /src/contrib/gcc/cp-demangle.c -o cp-demangle.So
/usr/local/bin/gcc33 -fpic -DPIC  -march=pentium2 -pipe -O2  -march=pentium2 -pipe -O2 -march=pentium2 -DIN_GLIBCPP_V3 -DHAVE_CONFIG_H -I/src/gnu/lib/libstdc++ -I/src/gnu/lib/libstdc++/../../../contrib/libstdc++/libsupc++ -I/src/gnu/lib/libstdc++/../../../contrib/gcc  -c /src/contrib/gcc/dyn-string.c -o dyn-string.So
building shared library libstdc++.so.4
/usr/obj/src/i386/usr/bin/ld: libstdc++.so.4: undefined versioned symbol name std::time_put_w@@GLIBCPP_3.2
/usr/obj/src/i386/usr/bin/ld: failed to set dynamic section sizes: Bad value
collect2: ld returned 1 exit status
*** Error code 1 (continuing)
`all' not remade because of errors.
===> gnu/lib/libsupc++
[...]

Also, for some reason, after installing this bad build, anything using /lib/libc.so.5 started having missing symbols unrelated to what is shown above, and I can't find a bad build of libc in this same buildworld log.  The log _shows_ libc.so.5 & its relatives all compiled and linked fine.  I did a rescue by copying libc.so.5 from a slightly old JPSNAP CD.

Doing a 'man' on anything now shows:
/libexec/ld-elf.so.1: /usr/bin/groff: Undefined symbol "__cxa_atexit"
or worse.

Could any of these glitches with gcc334 be a local makefile problem?
Or would they be bugs with gcc334 itself?

If it is bugs with gcc334, can someone update the lang/gcc33 port with a later snapshot, please?

Does anyone care that we try using a later gcc than what is installed with 5-current world?  "If it hurts don't do it"?  I feel sooner or later we will be forced into using it.

I guess I'll redo world with NO_DYNAMICROOT=yes and/or switch to the backlevel system gcc.

(Sorry for the bad formatting in this msg; I'm unable to send mail due to these problems and am cut-&-pasting into the html e-mail webform.)


  --  thx, Paul Seniura.

-- 
___________________________________________________________
Sign-up for Ads Free at Mail.com
http://promo.mail.com/adsfreejump.htm




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040408194909.8AAD1790058>