stlport fails to compile on my 22th June STABLE c++ -D_THREAD_SAFE -D_REENTRANT -Wall -g -O -I../../stlport -I. -DEH_VECTOR_OPERATOR_NEW nc_alloc.cpp -c -o obj/nc_alloc.o nc_alloc.cpp: In function `void * operator new(unsigned int)': nc_alloc.cpp:229: declaration of `operator new(unsigned int)' throws different exceptions /usr/include/g++/g++/new:28: previous declaration here nc_alloc.cpp: In function `void operator delete(void *)': nc_alloc.cpp:278: declaration of `operator delete(void *)' throws different exceptions /usr/include/g++/g++/new:30: previous declaration here gmake: *** [obj/nc_alloc.o] Error 1 *** Error code 2 Stop in /usr/ports/devel/stlport. root@kartoffel stlport> How-To-Repeat: update to latest stable cd /usr/ports/devel/stlport && make
Hi again, just FYI: the last commit doesn't fixed that problem. Greetings, Oliver -- :======> Oliver Lehmann <======: clear perl code is better than :====> lehmann@ans-netz.de <====: unclear awk code; but NOTHING :===> http://www.pofo.de/ <===: comes close to unclear perl code :=> http://wishlist.pofo.de/ <=: (taken from comp.lang.awk FAQ)
This is failing while doing a make test, which was added to the port=20 build without consulting me to demonstrate a bug in gcc. The main library is built with g++, while the test program is built=20 with c++. Can you verify that they're both pointing to the same=20 executable (which g++, which c++)? Do you have any aliases that=20 might conflict? The path to the new header file looks strange. I get=20 /usr/include/g++/new, not /usr/include/g++/g++/new. Do you have=20 multiple compilers installed? --=20 Paul Marquis pmarquis@pobox.com
On Tue, 25 Jun 2002 15:16:11 -0400 Paul Marquis <pmarquis@pobox.com> wrote: > The main library is built with g++, while the test program is built > with c++. Can you verify that they're both pointing to the same > executable (which g++, which c++)? it does > Do you have any aliases that > might conflict? nope > The path to the new header file looks strange. I get > /usr/include/g++/new, not /usr/include/g++/g++/new. Do you have > multiple compilers installed? nope olivleh1@kartoffel olivleh1> locate new | grep '/new$' /usr/include/g++/new /usr/local/include/stlport/new olivleh1@kartoffel olivleh1> ls -l /usr/include/g++/g++ lrwxr-xr-x 1 root wheel 16 Aug 21 2001 /usr/include/g++/g++ -> /usr/include/g++ olivleh1@kartoffel olivleh1> which g++ /usr/bin/g++ olivleh1@kartoffel olivleh1> which c++ /usr/bin/c++ olivleh1@kartoffel olivleh1> ls -l /usr/bin/g++ /usr/bin/c++ -r-xr-xr-x 3 root wheel 71376 Jun 23 01:43 /usr/bin/c++ -r-xr-xr-x 3 root wheel 71376 Jun 23 01:43 /usr/bin/g++ olivleh1@kartoffel olivleh1> ls -l /usr/include/g++/new -r--r--r-- 1 root wheel 1207 Nov 20 2000 /usr/include/g++/new olivleh1@kartoffel olivleh1> c++ --version 2.95.4 olivleh1@kartoffel olivleh1> g++ --version 2.95.4 olivleh1@kartoffel olivleh1> md5 /usr/bin/c++ MD5 (/usr/bin/c++) = 6557d12f2ce53f51163b80659036c1ff olivleh1@kartoffel olivleh1> md5 /usr/bin/g++ MD5 (/usr/bin/g++) = 6557d12f2ce53f51163b80659036c1ff olivleh1@kartoffel olivleh1> Greetings, Oliver -- :======> Oliver Lehmann <======: clear perl code is better than :====> lehmann@ans-netz.de <====: unclear awk code; but NOTHING :===> http://www.pofo.de/ <===: comes close to unclear perl code :=> http://wishlist.pofo.de/ <=: (taken from comp.lang.awk FAQ)
Another person had reported this problem a while back on -current and=20 I couldn't reproduce it, as I don't run current. We talked about=20 defining EH_DELETE_HAS_THROW_SPEC, but we didn't come to any=20 conclusions. I just upgraded to 4-STABLE and finally got the error. =20 It appears gcc 2.95.4 is more pedantic about exceptions than in=20 previous versions. I've verified that a slightly modified version of your patch works on=20 earlier versions of gcc, so I'm going to submit that. Hopefully, it=20 won't break -current. Eventually, I want to get rid of this portion of the port build, but=20 not before talking with the OpenOffice folks. --=20 Paul Marquis pmarquis@pobox.com
State Changed From-To: open->patched patch applied, plz cvsup and try it
State Changed From-To: patched->closed no longer a problem