From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.4.3DSL i686) I work for a company that is moving from sco to linux, we are running the 6.2 version with a 2.2.16-3-smp kernel and have a module called ibcs. This kernel is running the sco binaries just fine but we tried installing 7.1 on a test server, IBM 5600 with 2 processors, raid controller, 1 gig of ram, and one external drive array connected to the raid controller, but can only boot the smp kernel that comes with 7.1. Any attempt to load the enterprise kernel results in a kernel panic. I have used both initrd images and no initrd image at all but still get the panic. The server we want to put this on has 700 users telneted in and a website posting data as well. The main problems right now is whether or not the enterprise kernel is compiled with more than 256 pty's available, and are you going to build an rpm package for the linux abi module. We have tried to compile the most current abi module to the 2.4.2 kernel rpm source with no luck. The current smp kernel only has 256 pty's causing us to recompile the kernel. We have called support and paid for nothing, the person who was on your support staff did not have any info for us. Reproducible: Always Steps to Reproduce: 1.download 2.4.2 kernel rpm source 2.compile kernel with linux abi included - as a patch 3.boot the new kernel but the sco binaries will not run Actual Results: None of the sco binaries worked. Usually a segmentation fault occurs. Expected Results: Programs should have run properly. Processes should have been started during the boot process.
The enterprise kernel has support for >4gb of memory, and it uses a significant amount of memory to provide that support. You do not want to use it on a 1gb machine. The number of PTYs built in will not be different. Without debug information it is hard to guess why the enterprise kernel paniced. But for your case it is not so important because you do not want to use the enterprise kernel anyway -- more memory and cpu time overhead for no benefit. We will increase CONFIG_UNIX98_PTY_COUNT for future builds; thank you for bringing that need to our attention. We have marked the ibcs facility as deprecated for several releases. We'll listen for more customer feedback on this issue.
Will there be any support for the linux-abi?
We have marked the ibcs facility as deprecated for several releases. We'll listen for more customer feedback on this issue.
Oh, I see what the confusion was... For the 2.4 kernels, linux-abi replaces the ibcs package, but it was that functionality that we had marked as deprecated. So my point about waiting for more customer feedback on this issue really applies to linux-abi. Sorry for the confusion.
We do use iBCS for some parts of our originally SCO running apps, for the modules still not ported or impossible to port for link-libs problems; We built a stable rh6.2-based mix of emulated/compiled application installed at about 120 customers here in Italy... We'd like to start testing rh7.1 for all of the new installations (mainly for the new hardware support of 2.4 kernel), but if it lack in iBCS support... we can't use it... Can't you tell something more about future SCO emulation support?
sergio.tadini: The chances of future Red Hat kernels supporting it are VERY small. However, it might be that we provide a patch for linux-abi by some means, totally unsupported of course.
"Bob Meyers" <oregonbob2000> wrote in message news:zcwU6.73$5d.22958@newshog.newsread.com... > After reading the bugzilla report again I did see an email address for the > RedHat guy that stated: > > "We have marked the ibcs facility as deprecated for several releases. > We'll listen for more customer feedback on this issue." > > His email address is: johnsonm > > I intend to fire off a request for SCO binary capability. I encourage anyone > else who is interested to do so. It has been brought to the attention of myself and others like me that RedHat has officially decided not to bother continuing including or supporting IBCS in it's Linux distribution. I am lending my voice to say that the ability to run SCO binaries is a *MUST*. I have many customers currently on various versions of SCO osr5 and until I come across a linux distribution that is even 1/2 as dependable and problem-free as even the oldest, neglected, un-patched, version of any SCO OS has been, I still cannot recommend linux to new customers, even though now some are starting to ask for it specifically. There are many legacy apps out there that cannot be recompiled for a new platform, and cannot be replaced. and some actual current apps that are mission-critical, currently under development, and available for several flavors of unix, and even Windows, and *not* linux. The end result is, for those customers who need to upgrade to *something* in order to gain TCP/IP for instance, the options are the latest version of OpenServer even if it is a years-old stagnant product, or some version of linux that has a working ibcs. Currently this limits linux to 2.2.x kernels. I know about linux-abi. I have been tracking it's progress closely and have had it up & running ever since it first became available right up to and including the current 2.4.5 kernel. However, while it is working fine for *most* of my sco binaries that have no linux equivalent (after a little cosmetic hacking by myself), it is not working for at least one of my most important apps. There happens to be a linux-native version of that app, but it means re-purchasing the package (expensive!) and therefore eliminates linux as an upgrade path. (unless, as mentioned, you only upgrade to a 2.2.x linux) There are *tons* of current SCO cutomers out there, and SCO-loving consultants, who are starting to feel "forced" to look at linux against their will simply because there appears to be nothing that will be like SCO was, but at least linux will run their many and varied legacy, mission-critical, irreplaceable, custom applications. Non of us believes for a second that linux is a tenth as stable as SCO, and we (the consultants) and our customers value the dependability of the SCO OS above all else. And Our opinion of Linux (so far) and our opinion of SCO (the OS, not the company) was not misplaced! Even though our "evidence" is largely anecdotal, it is greatin quantity and consistant. With that in mind, I think any major Linux vendor who wants to sell to buisinesses, should think long and hard before dispensing with ibcs (or whatever new work-alike that may supercede it, such as linux-abi) Open Server 5.0.5 came out long enough ago that many who got it are ready to upgrade again. The OS is generally stable and reliable enough that an even larger currently installed base are still running much older versions, and they are doubly ready to upgrade, purely because their hardware is no longer up to current usage demands. Relatively few have seen a need to get the recently released 5.0.6, which seems to have very little in the way of improvements and updates that cannot be downloaded for free and applied to their existing 5.0.5 & 5.0.4 systems. That amounts to an *awful* lot of people basically looking at Linux. *Most* of them will either know already, or be advised by their consultants that they should not even consider any distribution that does not include a working ibcs so that they do not have to junk their prorietary software that in most cases they have been paying programmers to write and work on for many years. That is a very large investment that no one wants to throw away. Any distribution that tells them they should throw it away will leave a *very* bad taste in a *lot* of mouths. These are the kind of customers who paid $1500 for a *5 user* SCO license and then another for a 5 or 16 user filePro license, and then who knows how many thousands for some special purpose app, and then who knows how many thousands for subsequent custom programming. They will *pay* for a good product, and pay for support, even if they never actually use it, but the same thing that makes this so, also means none of them will tolerate a half-baked and problematical product. whichever linux distribution looks the most reliable, best supported, _and runs their sco binaries out of the box_ will get their business. pretty desktops and loads of gnome/kde multimedia toys, cutting edge kernel and application versions, and fancy automatic hardware detectors are not only passively unimportant to these customers, they are to some degree actively repellant. Brian K. White -- brian -- http://www.aljex.com/bkw/ +++++[>+++[>+++++>+++++++<<-]<-]>>+.>.+++++.+++++++.-.[>+<---]>++. filePro BBx Linux SCO Prosper/FACTS AutoCAD #callahans Satriani
If Linux wants to gain a stronger foothold in the commercial server arena, they should definitely include SCO Unix binary capability. SCO must still maintain one (if not the) of the largest installed base of commercial application servers. The fast moving development of Linux makes it a very attractive alternative and migration path for current SCO users. This idea is stifled however by the fact the Linux support for SCO binaries is waning. If you want to see how important it is, tune your news-reader to comp.unix.sco.misc, or do a google search for 'ibcs' or ABI in that group.
Our company (and our customers) is using "IBCS" to run SCO binary since 1996 We dont really understand why to throw away a good thing, without a valid substitute for it. We still have many customer, running on SCO and waiting an "upgrade" to linux, but without IBCS we still have to keep SCO UNIX. I Hope Linus and his guy want (or RedHat) come back or give us (the users) something good in change.
We have been been in the process of testing our software on some Redhat Linux systems for close to a year next to systems running on SCO boxes. The systems are up 24/7 and are used to process millions of records daily. We have been using ibcs and found we had to make only minor changes mainly because of incompatabilies of the Bash shell. It has been a long road to prove the stability of the Redhat boxes and were surprised to find they could process data faster than the SCO systems with the same hardware configuration. We have also replaced some of our W2K fileservers using Samba which has worked out very well. We have several machines out in the field, mainly Fortune 500 companies, running our software on SCO boxes also. Because of Redhat's decision to drop ibcs I am faced with a big problem not to mention having to deal with all the I told you so's that Linix was not for business. I picked Redhat because I believed they would go after the business market and hopefully would make the right decisions to do so. Now is Redhat's chance to do just that. Putting ibcs back in makes business sense leaving it out makes no sense.
I'd like to thank everyone who responded. We are experimenting more seriously with linux-abi and are seriously considering reversing our position on this. linux-abi probably won't be tier-1 supported, but will probably be supported just like ibcs was before, which seems to have been satisfactory in the past.
There is a current linux-abi module available now. We have tried it out and it works perfect on out 2.4.6 kernel. To get the latest linux-abi-2.4.6, go to this link: ftp.openlinux.org:/pub/people/hch/linux-abi/linux-abi-2.4.6.0.diff. All of our sco binaries are working perfectly under this version.
As you can see in rawhide (a snapshot of our development), a (currently slightly older, but that will get updated soon) version of this patch is in our current working kernel.
Hi SCO Binary support is essential for the next 3-> 5 years, there are many binaries that can *never* be ported to Linux (no source is available) and they must run on systems that replace existing SCO systems IMHO it was a daft decision to not include SCO Openserver binary support out of the box, in 15 years I have rarely seen a more stupid thing happen.
Kernel release 2.4.9-6 (released a few days ago) has the linux-abi patches included.