This is part of The Pile, a partial archive of some open source mailing lists and newsgroups.
Subject: RHL 7.0 From: Ivan Jager <ivanjager@bigfoot.com> Date: Sat, 30 Sep 2000 20:35:29 -0400 I finally finished downloading Red Hat Linux 7.0. I repartitioned my HD and did a complete install. During the installation one thing I noticed was that every time I pressed F1 it would say the help for that was not available. Is that only because I was doing an FTP install? I remember RHL 5.2 had help. The help for the packages was especially useful. I also noticed kernel 2.2.16. I had been expecting 2.4.*, but I guess Red Hat didn't want to wait. I played around a little, then decided to get something done. I tried to use MySQL, but /var/lib/mysql/ doesn't have the right permissions. I fixed that with a chmod +rx /var/lib/mysql/ . Why does Red Hat call the package "mysql" instead of "MySQL" like the RPMs from http://www.mysql.com/ ? That was ~3.5 hours after installing. Makes me wonder how much testing Red Hat does. The MySQL bug was already reported at bugzilla, and there were supposedly some updates, but the updates directory in ftp://ftp.redhat.com/pub/redhat/releases/guinness/ and in the mirrors is also lacking read permissions. (so I couldn't test them) (not to mention it is difficult to log in) === Subject: Re: RHL 7.0 From: Craig Kelley <ink@inconnu.isu.edu> Date: Tue, 3 Oct 2000 09:50:15 -0600 (MDT) On Sat, 30 Sep 2000, Ivan Jager wrote: > I finally finished downloading Red Hat Linux 7.0. I repartitioned my HD and did a complete install. > > During the installation one thing I noticed was that every time I > pressed F1 it would say the help for that was not available. Is that > only because I was doing an FTP install? I remember RHL 5.2 had help. > The help for the packages was especially useful. I noticed this as well. Some other observations: I upgraded my pinstripe which had kernel-2.2.16-17. RedHat 7.0 installed kernel-2.2.16-22 but anaconda did not run lilo (even though I remember saying 'linear' mode and all). So, I had to use the rescue disk to get it up (kudos to the new FTP rescue!). The fstab had been modified; there was only one kernel image in there where once there was three; but it looked like this: image=/boot/vmlinuz-2.2.16-16 label=linux read-only root=/dev/sda2 initrd=/boot/initrd-2.2.16-22.img strange, eh? A quick edit fixed the situation. === Subject: RE: RHL 7.0 From: "Barry Smoke" <barry@arhosting.com> Date: Tue, 3 Oct 2000 12:22:23 -0500 My installation did the same... Upon rebooting, I had my old kernel boot, which completely messed up all modules, edited lilo.conf, /sbin/lilo rebooted, depmod, couple more reboots, finally started working correctly. RH7.0 broke my x ...still haven't gotten it back up yet, I can't recompile the kernel... bugs..bugs..bugs.. === === Subject: about 7.0 and gcc 2.96 From: Levente Farkas <lfarkas@mindmaker.hu> Date: Thu, 12 Oct 2000 16:57:00 +0200 hi, first of all I don't know the overall quality of 7.0 since I'm just come back from my vacation, but I read many mails and news about in which people are criticize red hat why they include gcc 2.96 in 7.0. it seems to me in the last few years that there are just a few people who are using c++ on linux, but this seems more obvious after these responses. those who realy would like to use c++ first need a good c++ compiler. linux do not have any good (free) c++ compiler. egcs is not one of them. it's very limited on the filed of template, exception, namespace... those who realy argue have to look at more carefuly about it. gcc-2.96 which is not realy exists are much better, those who realy wanna use c++ with all of it feature would have be happy red hat finally include it in their distribution. if I were rh I even include libstdc++-v3 which is another essential part of a good c++ developer enviroment. now since v3 is the default stdc++ library in gcc I hope next rh distro will include a 'complete' standard c++ developmnet enviroment. ps. yes I know about the binary compatibility and the new ABI differencies, but than we can say rh have to include libc-5 and not glibc or glibc2.0, not glibc2.2 like now (glib2.1.9x is 2.2:-) and not the 2.4 kernel series... sometime have to change to the newer and it's high time now. when rh switch to glibc or elf from a.out than it was a good decision. sooner is better. === Subject: Re: about 7.0 and gcc 2.96 From: John Summerfield <summer@OS2.ami.com.au> Date: Fri, 13 Oct 2000 01:58:21 +0800 > first of all I don't know the overall quality of 7.0 since I'm just come > back from my vacation, but I read many mails and news about in which > people are criticize red hat why they include gcc 2.96 in 7.0. gcc 2.96 is a development version. Its authors say it's not ready for release. I don't think there would have been any argument had gcc 2.96 been put amongs the preview pacakges. However, RHI has chosen to build the entire distribution with it (kernel excepted) and has shipped two compilers anyway. What I find unsettling it the number of packages reshipped already; the fact so many problem packages have been found already suggests to me there are lots more to be found. Not just obscure packages that hardly anyone uses either; glibc is amongst them. > ps. yes I know about the binary compatibility and the new ABI differencies, > but than we can say rh have to include libc-5 and not glibc or glibc2.0, libc5 should indeed be included, but not as the base libc, but for applications users already have and which work well. Not everyone wants to go through the process of recompiling and retesting their applications just to move from RHL 5.x to 7.x. IN some cases, it's not an option; even Linux users use code that comes in binary-only. === Subject: Re: about 7.0 and gcc 2.96 From: jfm2@club-internet.fr Date: Thu, 12 Oct 2000 23:44:42 +0200 (CEST) > > first of all I don't know the overall quality of 7.0 > > since I'm just come back from my vacation, but I read > > many mails and news about in which people are criticize > > red hat why they include gcc 2.96 in 7.0. > gcc 2.96 is a development version. Its authors say it's > not ready for release. gcc 2.96 is a snapshot plus heavy patches. You could (or could not) consider that it is alike a production kernel where the code freeze has taken place before the official one by Linus. Not necessarily unstable but due to the fact this gcc is not official it has not got the same kind of post-release testing by users that an official gcc would have got. It only has got in house testing. This problem (of relative lack of testing) could have lightened if RH had anoounced in advance that this gcc would be the "next one". Given that it is not unusual that RedHat rolls back certain features of its betas I for one didn't pay much attention to the gcc in pinstripe and I didn't try to push it too hard. Also gcc 2.96 at the testsuite are far better than those of gcc 2.95. This could suggest that it is much better than people suggest. You could however wait a bit before using it for important things until other people have really tried it. > I don't think there would have been any argument had gcc > 2.96 been put amongs the preview pacakges. However, RHI > has chosen to build the entire distribution with it > (kernel excepted) and has shipped two compilers anyway. > What I find unsettling it the number of packages reshipped > already; the fact so many problem packages have been found > already suggests to me there are lots more to be > found. Not just obscure packages that hardly anyone uses > either; glibc is amongst them. This is not necessarily a gcc problem. You have to analyze the nature of the problems before saying it is. In 5.0 RH took a lot of flak for moving to glibc but when you look at the problems most of them were not related to it. It looked more like if the move to glibc had stretched RH resources to the point the building and testing of other programs was affected. === Subject: About 7.0 and gcc 2.96 From: Mario Torre <mario.torre@katamail.com> Date: Fri, 13 Oct 2000 01:39:22 +0200 I have finally installed the 7.0 and I have found it has al ot of bugs. Too many. Anaconda dies when I try to install files from both the CDs, I have updated the floppy iso from the update ftp directory, but I have I have not installed it again to test. I hope there is no need to reinstall everything, but, please, tell me if the basic installation compromise the file system. LPD doesn't find the road for the hostname if I use not the standard localhost. And, even if I don't change my host name, it doesn't seem to work (printer do not print). Pop3 doesn't works. I have installed a pop3 server to work with netscape for multiple e-mail, it worked on RedHat Linux 6.2, Slackware, Mandrake ... When called, the pop3 seems to die. Note that pop3 is configured correcly, it is not a bug of my pop3 daemon, I thing it is a bug in the tcp daemon, that is called by the pop3 program: >From the INSTALL file, for the inetd.conf: pop3 stream tcp nowait root /usr/sbin/tcpd popa3d RedHat L.7.0 uses xinetd, but the differences in configuration are not so difficult to understand. I use the popa3d-0.4 from openwall. Even if it cannot be attribuited to RedHat, the netscape pkg die when you chose ADDRESS in the Composer, and then click on a name and chose TO. RHL 6.2 had not this problem, Mandrake 7.0 had, but was fixed in the 7.1 release. While these problems are significants, they may be fixed by providing simple patches, so nothing very important - a little more beta testing is not so hard to do. And I think you are making beta testing with real users with the 7.0 release, so you can prepare yourself with the 8.0, that will be, probabily, your next release, with gcc3, kernel 2.4, kde2 and gnome revisited. I think I'm not so far from the truth... But it is not a bad thing, as who works with the RedHat Linux probabily have not upgrade to the 7.0, as it is not secure as it should be. But the worse problem, is that I cannot compile a kernel. I have used kgcc, I have edited the Makefile, used gcc, but nothing to do. Kernel compile errors somewhere, I have not take track of the bug, but I will do tomorrow, when I will try to re-compile the kernel with the gcc-2.95.2, as it is recommended. A notice, may be something that's about my system only, but the new GnoRPM should have a bug somewhere, cause it dies when I ask to install a new pkgs. I have upgraded to the new version, but it is worse than the previous one (it says seg-fault). Ah, something wrong there is also in the xemacs (it takes too much time to run...) and in some packages of the Powertools CD. But these are very minor problems, normal problems even of a great release. I think, anyway, that you should fix those bugs, above all the gcc/kgcc problem, the tcp and the lpd problem, and the netscape should not be bad... Hope to be useful to you all, I have enjoyed with RedHat, and I have used it, debug it, and I will use it in future, but I need to work, not to amuse myself with problems. Thanks to RedHat and to the Open Source, and good work. === Subject: Re: about 7.0 and gcc 2.96 From: Matt Wilson <msw@redhat.com> Date: Fri, 13 Oct 2000 01:05:32 -0400 On Fri, Oct 13, 2000 at 01:58:21AM +0800, John Summerfield wrote: > > What I find unsettling it the number of packages reshipped already; > the fact so many problem packages have been found already suggests > to me there are lots more to be found. Not just obscure packages > that hardly anyone uses either; glibc is amongst them. Note that none of these packages have been re-released because of compiler problems. The glibc fixes are mainly to repair broken backwards compatibility, not necessarily bugs in new functionality. > > ps. yes I know about the binary compatibility and the new ABI differencies, > > but than we can say rh have to include libc-5 and not glibc or glibc2.0, > > > libc5 should indeed be included, but not as the base libc, but for > applications users already have and which work well. Not everyone > wants to go through the process of recompiling and retesting their > applications just to move from RHL 5.x to 7.x. IN some cases, it's > not an option; even Linux users use code that comes in binary-only. I'll look into at least putting them on the FTP site in a conspicuous directory for users to download. === Subject: Re: about 7.0 and gcc 2.96 From: John Summerfield <summer@OS2.ami.com.au> Date: Fri, 13 Oct 2000 08:39:19 +0800 This is not necessarily a gcc problem. You have to analyze the nature > of the problems before saying it is. In 5.0 RH took a lot of flak for > moving to glibc but when you look at the problems most of them were > not related to it. It looked more like if the move to glibc had > stretched RH resources to the point the building and testing of other > programs was affected. I didn't intend anyone to think that I think gcc is a cause of any of the other problems; I see the number of problems reported and (in many cases fixed) already as an adverse reflection of the quality of RHL 7.0. gcc is only one of the problems. Dropping support for libc is another. btw Does anyone know about the new ISOs? What's changed? === Subject: Re: about 7.0 and gcc 2.96 From: John Summerfield <summer@OS2.ami.com.au> Date: Fri, 13 Oct 2000 15:16:48 +0800 libc5 should indeed be included, but not as the base libc, but for > > applications users already have and which work well. Not everyone > > wants to go through the process of recompiling and retesting their > > applications just to move from RHL 5.x to 7.x. IN some cases, it's > > not an option; even Linux users use code that comes in binary-only. > > I'll look into at least putting them on the FTP site in a conspicuous > directory for users to download. thanks Matt. You might put a README in the ISO director(y|ies) too. Not that anyone will read them;-), but you can say "We told you -- see ftp:.........." === Subject: Re: about 7.0 and gcc 2.96 From: Levente Farkas <lfarkas@mindmaker.hu> Date: Fri, 13 Oct 2000 11:22:00 +0200 jfm2@club-internet.fr wrote: > > > > > > hi, > > > first of all I don't know the overall quality of 7.0 since I'm just come > > > back from my vacation, but I read many mails and news about in which > > > people are criticize red hat why they include gcc 2.96 in 7.0. > > > > gcc 2.96 is a development version. Its authors say it's not ready for release. > > > > gcc 2.96 is a snapshot plus heavy patches. You could (or could not) > consider that it is alike a production kernel where the code freeze > has taken place before the official one by Linus. Not necessarily > unstable but due to the fact this gcc is not official it has not got > the same kind of post-release testing by users that an official gcc > would have got. It only has got in house testing. This problem (of > relative lack of testing) could have lightened if RH had anoounced in > advance that this gcc would be the "next one". Given that it is not > unusual that RedHat rolls back certain features of its betas I for one > didn't pay much attention to the gcc in pinstripe and I didn't try to > push it too hard. > > Also gcc 2.96 at the testsuite are far better than those of gcc 2.95. > This could suggest that it is much better than people suggest. You > could however wait a bit before using it for important things until > other people have really tried it. I've used 2.96 at least for 3 mounts from rawhide and use it with stdc++-v3 and it's much better then the earlier. may be not so many paople use it but it's time to test it by more people. when glibc comes and libc5 go away then there is just a few 'brave' people who use it and it was not a realy widely tested. for developers it's better, for users it's better if they don't have to know about it and they don't bother with it. does anybody have any real example why 2.96 is worse the 2.95 ? anyway 2.95 would be the worst choice, since it's worse some c++ part than egcs. ps. I read gcc's statement about it and I agree with them since they have to portect their 'product' from critic, and 2.96 is not their 'product'. and I've fully respect for them their great product. but from rh point of view there is many 'good' c++ compiler on windows, even ms's vc++ is better conforming to the standard c++ then the earlier gcc releases. until linux don't have a good c++ compiler windows c++ programmer can't turn to linux. there is a kai and will a borland c++ compiler for linux but more linux user would like to use gcc. === Subject: C++/STL compilation under 7.0 From: Steve Borho <steve@borho.org> Date: Wed, 4 Oct 2000 10:48:32 -0500 Has anyone tried to compile a C++ program which uses the stdc++ libs with 7.0? Initially the compile complains it can't find header files. A quick glance shows that /usr/include/g++ is no more and that /usr/include/g++-2 and -3 are now there... the -2 version is owned by the compat-libstc packages and the -3 are owned by the new compiler... Ok, I add /usr/include/g++-3 to my includes and start the compile again and I get a few hundred warning messages (complaining about stl_vector.h and template members) and it stops compiling. So I add /usr/include/g++-2 to the path instead and start the compile and it goes great until the linking stage where it barfs: ld: cannot find -lstdc++ Because there's two copies of the STL libraries installed: /usr/lib/libstdc++-2-libc6.1-1-2.9.0.a /usr/lib/libstdc++-2-libc6.1-1-2.9.0.so /usr/lib/libstdc++-libc6.1-1.a.2 -> libstdc++-2-libc6.1-1-2.9.0.a /usr/lib/libstdc++-libc6.1-1.so.2 -> libstdc++-2-libc6.1-1-2.9.0.so /usr/lib/libstdc++.so.2.7.2 -> libstdc++.so.2.7.2.8 /usr/lib/libstdc++.so.2.7.2.8 /usr/lib/libstdc++.so.2.8 -> libstdc++.so.2.8.0 /usr/lib/libstdc++.so.2.8.0 /usr/lib/libstdc++.so.2.9 -> libstdc++.so.2.9.dummy /usr/lib/libstdc++.so.2.9.dummy What's the proper way to make the compiler pick one or the other? Subject: C++/STL compilation under 7.0-- From: Steve Borho <steve@borho.org> Date: Wed, 4 Oct 2000 10:48:32 -0500 Has anyone tried to compile a C++ program which uses the stdc++ libs with 7.0? Initially the compile complains it can't find header files. A quick glance shows that /usr/include/g++ is no more and that /usr/include/g++-2 and -3 are now there... the -2 version is owned by the compat-libstc packages and the -3 are owned by the new compiler... Ok, I add /usr/include/g++-3 to my includes and start the compile again and I get a few hundred warning messages (complaining about stl_vector.h and template members) and it stops compiling. So I add /usr/include/g++-2 to the path instead and start the compile and it goes great until the linking stage where it barfs: ld: cannot find -lstdc++ Because there's two copies of the STL libraries installed: /usr/lib/libstdc++-2-libc6.1-1-2.9.0.a /usr/lib/libstdc++-2-libc6.1-1-2.9.0.so /usr/lib/libstdc++-libc6.1-1.a.2 -> libstdc++-2-libc6.1-1-2.9.0.a /usr/lib/libstdc++-libc6.1-1.so.2 -> libstdc++-2-libc6.1-1-2.9.0.so /usr/lib/libstdc++.so.2.7.2 -> libstdc++.so.2.7.2.8 /usr/lib/libstdc++.so.2.7.2.8 /usr/lib/libstdc++.so.2.8 -> libstdc++.so.2.8.0 /usr/lib/libstdc++.so.2.8.0 /usr/lib/libstdc++.so.2.9 -> libstdc++.so.2.9.dummy /usr/lib/libstdc++.so.2.9.dummy What's the proper way to make the compiler pick one or the other? === Subject: Re: C++/STL compilation under 7.0 From: Bernhard Rosenkraenzer <bero@redhat.de> Date: Wed, 4 Oct 2000 20:57:44 +0200 (CEST) On Wed, 4 Oct 2000, Steve Borho wrote: > Has anyone tried to compile a C++ program which uses the stdc++ libs > with 7.0? Sure - no problems at all here. > Initially the compile complains it can't find header files. A quick > glance shows that /usr/include/g++ is no more and that > /usr/include/g++-2 and -3 are now there... the -2 version is owned by > the compat-libstc packages and the -3 are owned by the new compiler... /usr/include/g++-3 is the correct one and is automatically used by the compiler. Are you sure you invoked the compiler as g++ and not just gcc? > ld: cannot find -lstdc++ > > Because there's two copies of the STL libraries installed: You should have /usr/lib/libstdc++-3-libc6.2-2-2.10.0.so Make sure you installed gcc-c++, libstdc++ and libstdc++-devel. Looks like at least one of them is missing on your system. === Subject: Re: C++/STL compilation under 7.0 From: Steve Borho <steve@borho.org> Date: Wed, 4 Oct 2000 16:57:11 -0500 On Wed, Oct 04, 2000 at 08:57:44PM +0200, Bernhard Rosenkraenzer wrote: > On Wed, 4 Oct 2000, Steve Borho wrote: > > > Has anyone tried to compile a C++ program which uses the stdc++ libs > > with 7.0? > > Sure - no problems at all here. > > > Initially the compile complains it can't find header files. A quick > > glance shows that /usr/include/g++ is no more and that > > /usr/include/g++-2 and -3 are now there... the -2 version is owned by > > the compat-libstc packages and the -3 are owned by the new compiler... > > /usr/include/g++-3 is the correct one and is automatically used by the > compiler. > Are you sure you invoked the compiler as g++ and not just gcc? ix% g++ -g -Wall -c -o pmcomp_input.o pmcomp_input.C In file included from pmcomp_input.C:16: pmcomp.hpp:21: iostream: No such file or directory pmcomp.hpp:22: vector: No such file or directory pmcomp.hpp:23: set: No such file or directory <snip> ix% g++ -g -Wall -I/usr/include/g++-3 -c -o pmcomp_input.o pmcomp_input.C In file included from /usr/include/g++-3/vector:31, from pmcomp.hpp:22, from pmcomp_input.C:16: /usr/include/g++-3/stl_alloc.h:742: invalid member template declaration /usr/include/g++-3/stl_alloc.h:780: invalid member template declaration /usr/include/g++-3/stl_alloc.h:817: invalid member template declaration /usr/include/g++-3/stl_alloc.h:858: invalid member template declaration /usr/include/g++-3/stl_alloc.h:961: empty component declaration /usr/include/g++-3/stl_alloc.h:961: parse error before `<' /usr/include/g++-3/stl_vector.h: In method `_Tp * vector<_Tp,_Alloc>::begin()': In file included from /usr/include/g++-3/vector:34, from pmcomp.hpp:22, from pmcomp_input.C:16: <snip> ix% g++ -g -Wall -I/usr/include/g++-2 -c -o pmcomp_input.o pmcomp_input.C ix% > You should have /usr/lib/libstdc++-3-libc6.2-2-2.10.0.so > > Make sure you installed gcc-c++, libstdc++ and libstdc++-devel. Looks like > at least one of them is missing on your system. ix% rpm -q gcc-c++ libstdc++ libstdc++-devel gcc-c++-2.96-54 libstdc++-2.96-54 libstdc++-devel-2.96-54 ix% which g++ /celox/bin/g++ Oooof. There it is. Sorry about that. I need to change my path. I'm getting the site wide compiler. Thanks for the help. ===