Re: X problems using egcs as compiler
* From: David O'Brien obr...@nuxi.com * So the fix we mentioned might not be needed with the base EGCS. Maybe * you can just compile all the ports just to see where we stand. I don't Well, as soon as the first snapshot with the system egcs is released, I'm planning to grab it and run the build through it. The logs will be at the normal place (http://bento/~asami/errorlogs/). -W To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
According to Alex Zepeda: Personally, I'd vote for using the new runtime objects, and forcing binary incompatibility. It's worth it IMO for the exception handling support if nothing else. However, if you're dead set against it, just back up your runtime objects, and edit the spec file (like the egcs port forced you to do at one time, and probably still does). The problem I see is that exceptions are for C++ and forcing ANSI C files to be compiled with -fexceptions and linked with new runtime objects is probably not the best way... -- Ollivier ROBERT -=- FreeBSD: The Power to Serve! -=- robe...@keltia.freenix.fr FreeBSD keltia.freenix.fr 4.0-CURRENT #70: Sat Feb 27 09:43:08 CET 1999 To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
According to David E . O'Brien: src/lib/csu/i386-elf will build and install crtbegin.o and crtend.o from the EGCS sources. Otherwise we have to use the poorer exception unwinding method. Is there a runtime overhead of using these files even for C programs ? -- Ollivier ROBERT -=- FreeBSD: The Power to Serve! -=- robe...@keltia.freenix.fr FreeBSD keltia.freenix.fr 4.0-CURRENT #70: Sat Feb 27 09:43:08 CET 1999 To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
In article 19990403193632.a55...@titan.klemm.gtn.com,
Andreas Klemm andr...@klemm.gtn.com wrote:
Got the latest egcs port
# $Id: Makefile,v 1.53 1999/03/30 02:58:02 obrien Exp $
Build X11R6 with the following CFLAGS:
-pipe -mpentiumpro -O2
I'm still running X11 and after ,make install' I'm unable
to launch x applications (xterm, ...)
andr...@titan{1001} $ xterm
/usr/libexec/ld-elf.so.1: /usr/X11R6/lib/libXaw.so.6: Undefined symbol
__deregister_frame_info
There are two things going on here. First, I believe you must have
built libXaw.so.6 using egcs, but you must have linked xterm using
the normal system compiler. (Or else you or the Makefile used ld
directly, which is always a bad idea.) If the executable is linked
using egcs, then this error won't happen.
Second, egcs actually contains a hack which attempts to make it work
anyway, even if you linked the program using the old compiler. But
there is a dynamic linker bug (which I just discovered) that prevents
the hack from working. I am testing the fix for that now.
John
--
John Polstra j...@polstra.com
John D. Polstra Co., Inc.Seattle, Washington USA
Self-interest is the aphrodisiac of belief. -- James V. DeLong
To Unsubscribe: send mail to majord...@freebsd.org
with unsubscribe freebsd-current in the body of the message
X problems using egcs as compiler
Got the latest egcs port
# $Id: Makefile,v 1.53 1999/03/30 02:58:02 obrien Exp $
Build X11R6 with the following CFLAGS:
-pipe -mpentiumpro -O2
I'm still running X11 and after ,make install' I'm unable
to launch x applications (xterm, ...)
andr...@titan{1001} $ xterm
/usr/libexec/ld-elf.so.1: /usr/X11R6/lib/libXaw.so.6: Undefined symbol
__deregister_frame_info
Exit 1
Is this a problem with egcs ? Or did I overlook something ?
My freebsd 3.1-STABLE kernel, compiled with -pipe -O2 -mpentiumpro,
runs fine since half a day.
--
Andreas Klemm http://www.FreeBSD.ORG/~andreas
http://www.freebsd.org/~fsmp/SMP/SMP.html
powered by Symmetric MultiProcessor FreeBSD
To Unsubscribe: send mail to majord...@freebsd.org
with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
David O'Brien forwarded me a message a couple of days ago that
contains a fix that may help. You need to add the following
line at line 301 in target.make
-u __deregister_frame_info \
-steve
On Sat, 3 Apr 1999, Andreas Klemm wrote:
# Got the latest egcs port
#
# # $Id: Makefile,v 1.53 1999/03/30 02:58:02 obrien Exp $
#
# Build X11R6 with the following CFLAGS:
#
# -pipe -mpentiumpro -O2
#
# I'm still running X11 and after ,make install' I'm unable
# to launch x applications (xterm, ...)
#
# andr...@titan{1001} $ xterm
# /usr/libexec/ld-elf.so.1: /usr/X11R6/lib/libXaw.so.6: Undefined symbol
__deregister_frame_info
# Exit 1
#
# Is this a problem with egcs ? Or did I overlook something ?
#
# My freebsd 3.1-STABLE kernel, compiled with -pipe -O2 -mpentiumpro,
# runs fine since half a day.
#
# --
# Andreas Klemm http://www.FreeBSD.ORG/~andreas
# http://www.freebsd.org/~fsmp/SMP/SMP.html
# powered by Symmetric MultiProcessor FreeBSD
#
#
# To Unsubscribe: send mail to majord...@freebsd.org
# with unsubscribe freebsd-current in the body of the message
#
To Unsubscribe: send mail to majord...@freebsd.org
with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Actually, it probably means that the program must be compiled with
-fexecptions.
Here is a little clue:
from egcs-1.1.2:
./gcc/frame.c:/* Called from crtbegin.o to deregister the unwind info for an
object. */
./gcc/frame.c:__deregister_frame_info (void *begin)
./gcc/frame.c:__deregister_frame (void *begin)
./gcc/frame.c: free (__deregister_frame_info (begin));
Just tried to compile apache and apache's jserv and got hit by the above bug
so I hope
that this gets resolved prior to using egcs as the default.
Amancio
David O'Brien forwarded me a message a couple of days ago that
contains a fix that may help. You need to add the following
line at line 301 in target.make
-u __deregister_frame_info \
-steve
On Sat, 3 Apr 1999, Andreas Klemm wrote:
# Got the latest egcs port
#
# # $Id: Makefile,v 1.53 1999/03/30 02:58:02 obrien Exp $
#
# Build X11R6 with the following CFLAGS:
#
# -pipe -mpentiumpro -O2
#
# I'm still running X11 and after ,make install' I'm unable
# to launch x applications (xterm, ...)
#
# andr...@titan{1001} $ xterm
# /usr/libexec/ld-elf.so.1: /usr/X11R6/lib/libXaw.so.6: Undefined symbol
__deregister_frame_info
# Exit 1
#
# Is this a problem with egcs ? Or did I overlook something ?
#
# My freebsd 3.1-STABLE kernel, compiled with -pipe -O2 -mpentiumpro,
# runs fine since half a day.
#
# --
# Andreas Klemm http://www.FreeBSD.ORG/~andreas
# http://www.freebsd.org/~fsmp/SMP/SMP.html
# powered by Symmetric MultiProcessor FreeBSD
#
#
# To Unsubscribe: send mail to majord...@freebsd.org
# with unsubscribe freebsd-current in the body of the message
#
To Unsubscribe: send mail to majord...@freebsd.org
with unsubscribe freebsd-current in the body of the message
To Unsubscribe: send mail to majord...@freebsd.org
with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Just tried to compile apache and apache's jserv and got hit by the above bug so I hope that this gets resolved prior to using egcs as the default. I doubt it will. I expect there will be significant breakage of non-world after the commit. I've significantly tested kernel and world, but I haven't tried ports at all. People may be well advised to CVSup and `make world' (or don't CVSup again for a while) sometime before Sunday California PDT time (-7hrs Zulu). The rockiness of -CURRENT is about to be upon us. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, David O'Brien wrote: # Just tried to compile apache and apache's jserv and got hit by the # above bug so I hope that this gets resolved prior to using egcs as the # default. # # I doubt it will. Me too. :/ # I expect there will be significant breakage of non-world after the # commit. I've significantly tested kernel and world, but I haven't tried # ports at all. I got everything setup to build the entire ports collection and sometime during the night the big IDE disk that I keep all the distfiles and packages on went south. :( It had been acting funny lately. I think it finally decided it had seen enough. I'm off to the store to pick up another and then by work where I have a T1 to grab all the distfiles that are not on the 3.1 CDs. I sure hope Satoshi is following this because it doesn't look like I'm going to be able to get very far before the switch is thrown. In fact now is probably a good time to post something to the ports mailing list. I'll send out something brief before I head off to the store. # People may be well advised to CVSup and `make world' (or don't CVSup # again for a while) sometime before Sunday California PDT time (-7hrs # Zulu). # # The rockiness of -CURRENT is about to be upon us. Committers brush up on your edit-pr skills, we're in for a load of fun in the following weeks. -steve # -- # -- David(obr...@nuxi.com -or- obr...@freebsd.org) # To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
According to Amancio Hasty: Actually, it probably means that the program must be compiled with -fexecptions. The problem is that egcs 1.1.2 has its own crt*.o files and use them in place of the FreeBSD one. By moving the files elsewhere, it works. I don't know the effect on exceptions though. I tumbled in this problem when compiling Perl 5.005_03 with egcs and moving the files away solved the problem. I don't know what David plan to do with these crt*.o files. David ? -- Ollivier ROBERT -=- FreeBSD: The Power to Serve! -=- robe...@keltia.freenix.fr FreeBSD keltia.freenix.fr 4.0-CURRENT #70: Sat Feb 27 09:43:08 CET 1999 To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, Ollivier Robert wrote: According to Amancio Hasty: Actually, it probably means that the program must be compiled with -fexecptions. The problem is that egcs 1.1.2 has its own crt*.o files and use them in place of the FreeBSD one. By moving the files elsewhere, it works. I don't know the effect on exceptions though. I tumbled in this problem when compiling Perl 5.005_03 with egcs and moving the files away solved the problem. I don't know what David plan to do with these crt*.o files. David ? Personally, I'd vote for using the new runtime objects, and forcing binary incompatibility. It's worth it IMO for the exception handling support if nothing else. However, if you're dead set against it, just back up your runtime objects, and edit the spec file (like the egcs port forced you to do at one time, and probably still does). - alex To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
gcc version egcs-2.91.66 19990314 (egcs-1.1.2 release) /usr/local/lib/gcc-lib/i386-portbld-freebsd3.1/egcs-2.91.66/collect2 -m elf_i38 6 -dynamic-linker /usr/libexec/ld-elf.so.1 -o resize /usr/lib/crt1.o /usr/lib/cr ti.o /usr/local/lib/gcc-lib/i386-portbld-freebsd3.1/egcs-2.91.66/crtbegin.o -L.. /../exports/lib -L/usr/X11R6/lib -L/usr/local/lib/gcc-lib/i386-portbld-freebsd3. 1/egcs-2.91.66 -L/usr/local/i386-portbld-freebsd3.1/lib -L/usr/local/lib resize. o -lxpg4 -lrpcsvc -lkrb -ldes -lcrypt -ltermcap -rpath /usr/X11R6/lib -lgcc -lc -lgcc /usr/local/lib/gcc-lib/i386-portbld-freebsd3.1/egcs-2.91.66/crtend.o /usr/ lib/crtn.o Looking at the link line, I don't see anything wrong on 1st glance. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
I got everything setup to build the entire ports collection and sometime during the night the big IDE disk that I keep all the I'm not sure the best way to approach this. The specification file used in the port is different from the one that will be in the base system. (the on in the Port is more stock EGCS) The base system will have many more of our hacks. So the fix we mentioned might not be needed with the base EGCS. Maybe you can just compile all the ports just to see where we stand. I don't expect *any* problems with ports that use the C compiler. All the problems I think we might have are C++ compiler issues. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
I have been compiling Netscape's Java VM Electric Fire which is written on C++ with no problems however I have been using the -fexceptions flag. Amancio To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Yes, it is most likely a DWARF2_UNWIND issue at least that is what I concluded in trying to support Electric Fire also someone should investigate whether egcs supports exception handling without thread support if memory does not fail me egcs requires thread support for exception handling. Perhaps it is a good time to review the egcs docs with respect to frame handling and exception handling. Amancio The problem is that egcs 1.1.2 has its own crt*.o files and use them in place of the FreeBSD one. By moving the files elsewhere, it works. I don't know the effect on exceptions though. If that is the problem, then it sounds like a DWARF2_UNWIND issue. I don't know what David plan to do with these crt*.o files. David ? src/lib/csu/i386-elf will build and install crtbegin.o and crtend.o from the EGCS sources. Otherwise we have to use the poorer exception unwinding method. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, Amancio Hasty wrote: Yes, it is most likely a DWARF2_UNWIND issue at least that is what I concluded in trying to support Electric Fire also someone should investigate whether egcs supports exception handling without thread support if memory does not fail me egcs requires thread support for exception handling. Perhaps it is a good time to review the egcs docs with respect to frame handling and exception handling. How can I tell if I've got thread support enabled? I've used ksirc before which uses exceptions to parse irc output, and it's obviously not threaded (cause Qt isn't). Sure, I don't like ksirc, and I got some odd perl related errors, but it worked. - alex To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, David O'Brien wrote: # I'm not sure the best way to approach this. The specification file used # in the port is different from the one that will be in the base system. # (the on in the Port is more stock EGCS) The base system will have many # more of our hacks. I don't know what we'd prove if the ports tree was built with an un-FreeBSD'd egcs. It might get us close, then again it might get us chasing (mis)features that aren't going to be present when it is the system compiler. I certainly would feel better about using the real thing. # So the fix we mentioned might not be needed with the base EGCS. Maybe # you can just compile all the ports just to see where we stand. I don't # expect *any* problems with ports that use the C compiler. All the # problems I think we might have are C++ compiler issues. See above. Can you send me a set of instructions for turning gcc off and egcs on in the base distribution? -steve # -- # -- David(obr...@nuxi.com -or- obr...@freebsd.org) # To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
See above. Can you send me a set of instructions for turning gcc off and egcs on in the base distribution? Better yet, can we just run with this today instead of tomorrow? I see no reason to wait any further, and some of us have more time this weekend to work on testing things which a Sunday schedule would collapse to less than half a day. - Jordan To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Better yet, can we just run with this today instead of tomorrow? Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make world'. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, Jordan K. Hubbard wrote: # Better yet, can we just run with this today instead of tomorrow? I # see no reason to wait any further, and some of us have more time this # weekend to work on testing things which a Sunday schedule would # collapse to less than half a day. You're right, even better yet. Throw the switch already! :) -steve To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
David O'Brien wrote: Better yet, can we just run with this today instead of tomorrow? Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make world'. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message Just did that... results in roughly 1 hour 1/2, ok? (unless it breaks, of course). PYD To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
On Sat, 3 Apr 1999, David O'Brien wrote: # Better yet, can we just run with this today instead of tomorrow? # # Can someone with a FAST machine do a CVSup from Freefall and `make world' # to verify that it is not broken? Doing it now. Going out to grab a bite to eat with the family but I should be back before it finishes. It usually only takes a little over an hour to do a 'make world' from scratch on my box at work. I'll let you know when it gets done. # My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make # world'. # # -- # -- David(obr...@nuxi.com -or- obr...@freebsd.org) # To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
At 04:28 PM 4/3/99 -0800, David O'Brien wrote: Better yet, can we just run with this today instead of tomorrow? Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make world'. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message I started a Make world about 1 hour ago with current as of then sources. It should be done in about 30min's. The only thing is I don't build xntpd but every thing else is built. I'll post when completed Manfred = ||man...@pacbell.net || ||Ph. (415) 681-6235|| = To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
David O'Brien wrote: Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? CVSupped from 01:25 BST, make -j12 -DCLOBBER world on an SMP box, both ELF and aout libraries generated, compile options -02 -m486, all clear. The word is go. PYD To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Just finished doing a cvsup and currently doing a make world. Should be done in less than 80 minutes . Cheers, Amancio To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Once you say the word that egcs is the default and a make world will actually use it, I'll be more than happy to do so immediately with my dual PII-450 box here. - Jordan Better yet, can we just run with this today instead of tomorrow? Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make world'. -- -- David(obr...@nuxi.com -or- obr...@freebsd.org) To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
-Original Message- From: David O'Brien obr...@nuxi.com To: Jordan K. Hubbard j...@zippy.cdrom.com; Steve Price spr...@hiwaay.net Cc: curr...@freebsd.org curr...@freebsd.org Date: Saturday, April 03, 1999 19:29 Subject: Re: X problems using egcs as compiler Better yet, can we just run with this today instead of tomorrow? Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? My machines are either EGCS'ified, 3.1-CURRENT, or take 5 hours to `make world'. Done. cvsupped at 19:48 EST from cvsup2.freebsd.org buildworld with stock make.conf just finished without any problems. I'd say go for it :) Steph --- Stephane E. Potvin Galea Network Security To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
At 4:28 pm -0800 3/4/99, David O'Brien wrote: Can someone with a FAST machine do a CVSup from Freefall and `make world' to verify that it is not broken? elf make world completed on Sun Apr 4 03:07:53 BST 1999 cvsup'd (from the UK server) at Sun Apr 4 01:50:56 BST 1999 -- Bob Bishop (0118) 977 4017 international code +44 118 r...@gid.co.ukfax (0118) 989 4254 between 0800 and 1800 UK To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
cvsup around 17:30 PST -- elf make world started on Sat Apr 3 17:33:29 PST 1999 -- -- elf make world completed on Sat Apr 3 18:33:39 PST 1999 -- Amancio To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
Re: X problems using egcs as compiler
Only thing I can see to note is that the cvs mailing lists are still apparently down, so notification of this (and any other changes which take place) won't go out to the interested listeners. If jmb is planning to re-send all of the bounced email when he fixes the problem, that's okay. Kris - The Feynman problem-solving algorithm: 1. Write down the problem 2. Think real hard 3. Write down the solution To Unsubscribe: send mail to majord...@freebsd.org with unsubscribe freebsd-current in the body of the message
