Re: Compiling 4-RELEASE on 5-STABLE
On Thu, 14 Oct 2004 01:18:25 -0400 (EDT), John Gillis [EMAIL PROTECTED] wrote: however.. there are still things that could go wrong and I prefer not to find out when a production server heads south or doesn't have the right firmware on the RAID card.. so that's why I lag. You can very well lag but doing so you would be creating apain in your forehead. Because everything that worked on 4.X-Release ( repeat, notice the word release) will also work on 5.Y-Release ( notice the word release). And also I would argue against keeping your boxes in the 4.X tree because as new releases are made available, the source codes of the ports are also changed to make it portable to the new tree. Most of the times tyhey DO compile on the old tree but you would be loosing features and performance. For example, there is a huge difference in performance when Xorg is compiled on gcc 2.95 and on gcc 3.4.2. If you are interested then contact me off the list. I would be happy to send you the performance monitor logs. Also, at least one piece of hardware is near impossible to upgrade. An old 486/25 that's running Snort, without a cd-rom and a 200M hard drive. Negative, this hardware is also upgradable. Just the catch is it is not SELF upgradable. I mean you can expect to compile the tree on the 486 itself. But you can very well compile its tree and kernel on a Pentium (say) and install it on the 200M harddisk. I have quite a few routers which have a lot similar configuration to yours and they work till date without complaining. Just the upgradng is a bit troublesome compared to the newer systems. Refer to the CFLAGS parameter in man make.conf What about my question about boot strapping? Does that ensure that I could compile the world/kernel of 4.x on 5.3? John No it does not. To be precise as far as I know, there is no way you can compile a native 4.X binary of any kind (application, kernel, bootstrap, you name it) on a 5.X box. Although you CAN run native 4.X binaries on a 5.X kernel using the compatibility layers. Refer to /usr/src/sys/NOTES for further information about compiling the compatibility layers. And yes the boot strap code has also changed from 4.X. Refer to /usr/src/UPDATING for a brief log about what had been changed since the 4.X tree. Also you can find detailed information about upgrading a 4.X tree to a 5.Y tree (although not recommended for other performance, security and a number of other reasons). Regards S. -- Subhro Sankha Kar School of Information Technology Block AQ-13/1 Sector V ZIP 700091 India ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 12:31:33PM +0530, Subhro wrote: What about my question about boot strapping? Does that ensure that I could compile the world/kernel of 4.x on 5.3? John No it does not. To be precise as far as I know, there is no way you can compile a native 4.X binary of any kind (application, kernel, bootstrap, you name it) on a 5.X box. Although you CAN run native 4.X [...] As a side note, with NetBSD you CAN do this. You can even compile NetBSD/vax on a Linux/ppc box. :-) That's how they build releases for their obscurer architectures all the time. GH -- :wq ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Wed, 2004-10-13 at 18:11, John Gillis wrote: My apologies if this has already been asked. I'd like to upgrade my non-production machines to 5.3 once it is released, however I'd like the production servers to lag behind once I make sure everything is working right. This might mean that my production servers would be running 4.x for the next few months. Compiling world, the kernel, and ports is done on non-production machines however, with the ports being packaged and installed on the servers and /usr/src being NFS mounted from a non-production machine. After installing 5.3 on the non-production machines, I'd like to track the 4-RELEASEs into another directory, say /usr/src.4 while tracking 5.3-RELEASE in /usr/src. My question is.. would I be able to compile anything on 5.3 that would still work on 4.10? Does the make build(world|kernel) bootstrap and then use the /usr/src.4 development environment to link/compile everything? Would I be at a loss with ports (not terribly important in my environment)? Consider holding back one non-production machine at 4.x as the build box for the production servers. That would eliminate any chance of the 5.x tree affecting the production servers. Either run cvsup twice (once from the 4.x build box and again from the 5.x build box) or set up your own cvsup mirror. -- Gary Dunn [EMAIL PROTECTED] Honolulu ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 12:31:33PM +0530, Subhro wrote: On Thu, 14 Oct 2004 01:18:25 -0400 (EDT), John Gillis [EMAIL PROTECTED] wrote: however.. there are still things that could go wrong and I prefer not to find out when a production server heads south or doesn't have the right firmware on the RAID card.. so that's why I lag. You can very well lag but doing so you would be creating apain in your forehead. Because everything that worked on 4.X-Release ( repeat, notice the word release) will also work on 5.Y-Release ( notice the word release). If you actually believe that I have a very nice bridge here you might be interested in. It is certainly the goal that things which worked on 4.x will continue to work in 5.x, and it might even work out that way in 99.99% of all cases, but *everything*? Not a bloody chance - there are always bugs that have yet to fixed (or even discovered). Also, at least one piece of hardware is near impossible to upgrade. An old 486/25 that's running Snort, without a cd-rom and a 200M hard drive. Negative, this hardware is also upgradable. Depends. It it is a 486sx it will not run 5.x (support for FPU-less systems has been removed.) I believe 5.x also needs a bit more memory than 4.x, so if that box has too little RAM it might be unbearably slow under 5.x What about my question about boot strapping? Does that ensure that I could compile the world/kernel of 4.x on 5.3? John No it does not. To be precise as far as I know, there is no way you can compile a native 4.X binary of any kind (application, kernel, bootstrap, you name it) on a 5.X box. And that is bullshit. It is of course possible to compile a 4.x binary on a 5.x box - just make sure you link against the right libraries (and in the case of C++ programs at least, compile with a compatible compiler.) I don't know if it is possible to do this without jumping through an inordinate number of hoops however. -- Insert your favourite quote here. Erik Trulsson [EMAIL PROTECTED] ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 09:52:52AM +0530, Subhro wrote: On Thu, 14 Oct 2004 00:11:05 -0400 (EDT), John Gillis [EMAIL PROTECTED] wrote: My apologies if this has already been asked. I'd like to upgrade my non-production machines to 5.3 Nice idea once it is released, however I'd like the production servers to lag behind once I make sure everything is working right. If everything is not working right, then 5.3 wouldnever be tagged STABLE. This is not Windows. 5.3 will have bugs even when tagged -stable. I can guarantee that once 5.3 is released and people start using it more widely there will be new problems being reported that haven't been discovered yet. 4.10-R uses gcc 2.95 and 5.3 uses gcc 3.4. The binaries compiled with the later are not backward compatble. You do realise that you can install gcc 3.4 on a 4.x machine and run the binaries compiled with it? For C++ the ABI has changed a couple of times betwenn gcc 2.95 and gcc 3.4, but for C everything should work fine, as long as you link against the right libraries. -- Insert your favourite quote here. Erik Trulsson [EMAIL PROTECTED] ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, 14 Oct 2004 12:01:35 +0200, Erik Trulsson [EMAIL PROTECTED] wrote: If you actually believe that I have a very nice bridge here you might be interested in. It is certainly the goal that things which worked on 4.x will continue to work in 5.x, and it might even work out that way in 99.99% of all cases, but *everything*? Not a bloody chance - there are always bugs that have yet to fixed (or even discovered). The primary phrase which a developer must believe is No software is 100% foolproof. So speaking in that line, indeed no software including the Releases of FreeBSD are 100% bug free. What I meant was, the RELEASE, not STABLE is expected to work correctly and fight back all the bugs that had been discovered till date. But I never meant it is perfect. If it was, then we would never have patches or future releases. And BTW I would be really interested to know about some hardware/software which used to work under 4.X and stopped working under 5.Y even after updating to the latest versions and applying all patches/hacks. It is entirely probable and acceptable that out of the box, a software natively made for 4.X will not work on 5.Y Depends. It it is a 486sx it will not run 5.x (support for FPU-less systems has been removed.) I believe 5.x also needs a bit more memory than 4.x, so if that box has too little RAM it might be unbearably slow under 5.x Yeh, I forgot to mention about the FPU. Thanks for adding up. And that is bullshit. It is of course possible to compile a 4.x binary on a 5.x box - just make sure you link against the right libraries (and in the case of C++ programs at least, compile with a compatible compiler.) I don't know if it is possible to do this without jumping through an inordinate number of hoops however. First of all, I guess u got a bit too aggresive which I believe is unnecessary. Secondly, Try disassembling a 4.X binary and a 5.X binary, you will understand what I mean. I have done it myself and I am sure about it. Things start differing even more when you start enabling things like unrolling loops and making things architecture dependant with mcpu, march and similar flags. And btw I guess you missed a word in my previous mail. I added the word native. Do clarify if that was not clear to you what I meant by Native. Regards S. -- Subhro Sankha Kar School of Information Technology Block AQ-13/1 Sector V ZIP 700091 India ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, 14 Oct 2004 12:06:19 +0200, Erik Trulsson [EMAIL PROTECTED] wrote: snip You do realise that you can install gcc 3.4 on a 4.x machine and run the binaries compiled with it? For C++ the ABI has changed a couple of times betwenn gcc 2.95 and gcc 3.4, but for C everything should work fine, as long as you link against the right libraries. What I meant was the system compiler. And I have tried to use the compiler from the ports as the system compiler. Try it out yourself and you will know what I mean. Regards S. -- Subhro Sankha Kar School of Information Technology Block AQ-13/1 Sector V ZIP 700091 India ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
RE: Compiling 4-RELEASE on 5-STABLE
-Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of Subhro Sent: Thursday, October 14, 2004 2:24 PM To: John Gillis; FreeBSD Questions Subject: Re: Compiling 4-RELEASE on 5-STABLE Depends. It it is a 486sx it will not run 5.x (support for FPU-less systems has been removed.) I believe 5.x also needs a bit more memory than 4.x, so if that box has too little RAM it might be unbearably slow under 5.x Yeh, I forgot to mention about the FPU. Thanks for adding up. Just on a side note here, could someone explain to me what a 'FPU-less' system actually is? Regards -- Mick Walker NAAFI Finance International ** This email and any files transmitted with it are confidential, and may be subject to legal privilege, and are intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error or think you may have done so, you may not peruse, use, disseminate, distribute or copy this message. Please notify the sender immediately and delete the original e-mail from your system. Computer viruses can be transmitted by e-mail. Recipients should check this e-mail for the presence of viruses. The Capita Group and its subsidiaries accept no liability for any damage caused by any virus transmitted by this e-mail. *** ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 02:45:38PM +0100, Walker, Michael wrote: [mailto:[EMAIL PROTECTED] Behalf Of Subhro Yeh, I forgot to mention about the FPU. Thanks for adding up. Just on a side note here, could someone explain to me what a 'FPU-less' system actually is? Ancient history. FPU means 'Floating Point Unit' -- ie. the bit of the CPU that handles arithmetic etc. on floading point numbers. Way back in the mists of processor pre-history, the original 8086 was an integer-only CPU. The same applied to the 80186, 80286 and 80386, although by that time it had grown a companion chip the 80n87 which held the FPU. By the time the 80486 came along, the FPU was incorporated into the main CPU silicon, although there were some cheapo 486 chips where the FPU had failed during manufacturing, sold as integer only processors. In order to run Unix on such systems, it was necessary to provide a software library to emulate the FPU system. FreeBSD actually came with a choice of two. All obsolete now. All other IA32 architecture CPUs have had a built in FPU as standard, as do all modern general purpose CPU chips. Cheers, Matthew -- Dr Matthew J Seaman MA, D.Phil. 26 The Paddocks Savill Way PGP: http://www.infracaninophile.co.uk/pgpkey Marlow Tel: +44 1628 476614 Bucks., SL7 1TH UK pgpxcIcQd3FXm.pgp Description: PGP signature
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 06:53:31PM +0530, Subhro typed: [...] First of all, I guess u got a bit too aggresive which I believe is unnecessary. Secondly, Try disassembling a 4.X binary and a 5.X binary, you will understand what I mean. I have done it myself and I am sure about it. Things start differing even more when you start enabling things like unrolling loops and making things architecture dependant with mcpu, march and similar flags. And btw I guess you missed a word in my previous mail. I added the word native. Do clarify if that was not clear to you what I meant by Native. Install 4.x in a subdirectory of your 5.x system. chroot(8) into that directory; then build native world and kernels all you want. too easy. Ruben ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, Oct 14, 2004 at 12:31:33PM +0530, Subhro wrote: What about my question about boot strapping? Does that ensure that I could compile the world/kernel of 4.x on 5.3? John No it does not. To be precise as far as I know, there is no way you can compile a native 4.X binary of any kind (application, kernel, bootstrap, you name it) on a 5.X box. It's not so hard, you just extract a 4.x system image into a directory and chroot to it, and build from there. With more work you could do it without the chroot. However, to answer the original question: binaries built by the 5.x system toolchain cannot be run on 4.x. The main barrier is that 5.x binaries usually use syscalls that are not present in the 4.x kernel, so you'll get signal 12 errors if you try and run them. Kris pgpWaPYKpgiG8.pgp Description: PGP signature
Re: Compiling 4-RELEASE on 5-STABLE
On Thu, 14 Oct 2004 00:11:05 -0400 (EDT), John Gillis [EMAIL PROTECTED] wrote: My apologies if this has already been asked. I'd like to upgrade my non-production machines to 5.3 Nice idea once it is released, however I'd like the production servers to lag behind once I make sure everything is working right. If everything is not working right, then 5.3 wouldnever be tagged STABLE. This is not Windows. This might mean that my production servers would be running 4.x for the next few months. Compiling world, the kernel, and ports is done on non-production machines however, with the ports being packaged and installed on the servers and /usr/src being NFS mounted from a non-production machine. 4.10-R uses gcc 2.95 and 5.3 uses gcc 3.4. The binaries compiled with the later are not backward compatble. After installing 5.3 on the non-production machines, I'd like to track the 4-RELEASEs into another directory, say /usr/src.4 while tracking 5.3-RELEASE in /usr/src. This can be done without trouble. man cvsup. snip Regards S. -- Subhro Sankha Kar School of Information Technology Block AQ-13/1 Sector V ZIP 700091 India ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: Compiling 4-RELEASE on 5-STABLE
If everything is not working right, then 5.3 wouldnever be tagged STABLE. This is not Windows. Granted, however I need to make sure that everything works fine in my environment on my hardware with my software. The reason I use FreeBSD is that I trust your release engineering moreso than any other vender (as it were)*, however.. there are still things that could go wrong and I prefer not to find out when a production server heads south or doesn't have the right firmware on the RAID card.. so that's why I lag. Also, at least one piece of hardware is near impossible to upgrade. An old 486/25 that's running Snort, without a cd-rom and a 200M hard drive. 4.10-R uses gcc 2.95 and 5.3 uses gcc 3.4. The binaries compiled with the later are not backward compatble. What about my question about boot strapping? Does that ensure that I could compile the world/kernel of 4.x on 5.3? John * Are you a vendor? You don't vend... I should say that I trust you more than any other operating system.. but that's too wordy. ___ [EMAIL PROTECTED] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to [EMAIL PROTECTED]