From: UnixOS2 Archive To: "UnixOS2 Archive" Date: Fri, 25 Jan 2002 04:09:51 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 115 ************************************************** Thursday 24 January 2002 Number 115 ************************************************** Subjects for today 1 Re: LIBEMU ? : Jack Troughton 2 Re: LIBEMU ? : Jack Troughton 3 Re: LIBEMU ? : Jack Troughton 4 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : Adrian Gschwend" 5 Re: LIBEMU ? : John Poltorak 6 Re: LIBEMU ? : Holger Veit 7 Re: LIBEMU ? : Holger Veit 8 Re: LIBEMU ? : John Poltorak 9 BSD stack (was Re: LIBEMU ?) : Adrian Gschwend" 10 Re: LIBEMU ? : John Poltorak 11 Re: BSD stack (was Re: LIBEMU ?) : Holger Veit 12 Re: BSD stack (was Re: LIBEMU ?) : John Poltorak 13 Re: LIBEMU ? : Holger Veit 14 Re: LIBEMU ? : Adrian Gschwend" 15 Re: LIBEMU ? : Holger Veit 16 Re: netCDF : Christian Hennecke" 17 Re: BSD stack (was Re: LIBEMU ?) : Adrian Gschwend" 18 Re: LIBEMU ? : Holger Veit 19 Re: LIBEMU ? : Holger Veit 20 Re: LIBEMU ? : Holger Veit 21 Autoconf 2.52g is released : John Poltorak 22 Re: LIBEMU ? : Gerhard Arnecke" 23 Re: vxcron (was Re: Cron/2 1.41 DOS Session error? : Andrew Belov" 24 Error with Perl 561 : Tim Erickson" 25 Re: Error with Perl 561 : Henry Sobotka **= Email 1 ==========================** Date: Fri, 25 Jan 2002 06:25:56 -0500 From: Jack Troughton Subject: Re: LIBEMU ? Holger Veit wrote: > > On Fri, Jan 25, 2002 at 10:22:38AM +0000, John Poltorak wrote: > > On Fri, Jan 25, 2002 at 10:59:49AM +0100, Holger Veit wrote: > > > That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with > > > Software Choice, MCP/ACP and eCS (so *I* have it), and that it once > > > in a while leaks through testcase. Do you have a definite URL for > > > downloading, so that we can fix this together with the the >FP13 > > > requirement. This will quite simplify the whole procedure. > > > > A complete refresh of MPTS is available here:- > > > > ftp://service.boulder.ibm.com/ps/products/mpts/fixes/english-us/wr08610_conv/ > > > > Here's what README.TXT says:- > > > > This FixPak is a refresh of the MPTS LAN Adapter and Protocol Support > > program and is intended to replace (upgrade) only LICENSED instances > > of MPTS. > > This is not TCP/IP 4.3. > > > > My impression was that meanwhile certain fixes for drivers (and maybe even > > > this V4.3 upgrade) are only available if you buy certain support. > > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > stack? > > The complete package which you can download from software choice if you > have a license. Socket support is in the MPTS refresh that John mentions above. Most of the changes between 4.1 and 4.3 is not in the stack, but in the applications included with the stack; for example, in 4.3 one gets dhcp and ddns servers. This is evidenced by the fact that whether or not one has 4.1 or 4.3, you have to apply MPTN fix IC27649 in order to get rid of the "trap after reconnecting a ppp connection" problem. You should be able to tell people to get 8620 and then ic27649; it is legal and should give you the services you need. It's the applications that IBM is protecting, not the stack itself. Regards, Jack **= Email 2 ==========================** Date: Fri, 25 Jan 2002 06:31:47 -0500 From: Jack Troughton Subject: Re: LIBEMU ? Holger Veit wrote: > > On Fri, Jan 25, 2002 at 10:45:44AM +0000, John Poltorak wrote: > > > This is not TCP/IP 4.3. > > > > Yes, I know. It's the MPTS stack which was available in TCP/IP 4.1 which > > was originally only available through Software Choice. > > > > > > > > > My impression was that meanwhile certain fixes for drivers (and maybe even > > > > > this V4.3 upgrade) are only available if you buy certain support. > > > > > > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > > > stack? > > > > > > The complete package which you can download from software choice if you > > > have a license. > > > > My question is related to whether the v4.3 upgrade is an essential > > requirement or whether the features available in 4.1 are sufficient. I'm > > not really sure what the real differences are. > > I was under the impression that the 4.x (x > 0) version contains DLLs > which export entries like socketpair() and which have improved AF_UNIX > support. This is not in so32dll.dll, but in a different DLL. I just > don't know whether x=1 or x=3 is required. Hmmmm.... [C:\mptn\dll]exeinfo -x tcpip32.dll | grep -i pair ExeInfo, New Exe File Format Info, v2.0f (C) Copyright Ngb Technologies, 1991-1993, All rights reserved. 254 SOCKETPAIR I think it's in tcpip32.dll :) Regards, Jack **= Email 3 ==========================** Date: Fri, 25 Jan 2002 06:37:43 -0500 From: Jack Troughton Subject: Re: LIBEMU ? Jack Troughton wrote: > > Holger Veit wrote: > > > > On Fri, Jan 25, 2002 at 10:45:44AM +0000, John Poltorak wrote: > > > > This is not TCP/IP 4.3. > > > > > > Yes, I know. It's the MPTS stack which was available in TCP/IP 4.1 which > > > was originally only available through Software Choice. > > > > > > > > > > > > My impression was that meanwhile certain fixes for drivers (and maybe even > > > > > > this V4.3 upgrade) are only available if you buy certain support. > > > > > > > > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > > > > stack? > > > > > > > > The complete package which you can download from software choice if you > > > > have a license. > > > > > > My question is related to whether the v4.3 upgrade is an essential > > > requirement or whether the features available in 4.1 are sufficient. I'm > > > not really sure what the real differences are. > > > > I was under the impression that the 4.x (x > 0) version contains DLLs > > which export entries like socketpair() and which have improved AF_UNIX > > support. This is not in so32dll.dll, but in a different DLL. I just > > don't know whether x=1 or x=3 is required. > > Hmmmm.... > > [C:\mptn\dll]exeinfo -x tcpip32.dll | grep -i pair > > ExeInfo, New Exe File Format Info, v2.0f > (C) Copyright Ngb Technologies, 1991-1993, All rights reserved. > > 254 SOCKETPAIR > > I think it's in tcpip32.dll :) One note about this... this system has IC27649 on it (necessary if you don't want afinetk to trap). You can (and definitely should!) apply this fix to 4.1... and tcpip32.dll is in the fix. Regards, Jack **= Email 4 ==========================** Date: Fri, 25 Jan 2002 08:57:57 +0100 (CET) From: "Adrian Gschwend" Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Thu, 24 Jan 2002 02:51:37 +0300 (MSK), Andrew Belov wrote: >CRONTAB.C tolerates both %VISUAL% and %EDITOR%, so try both. If %EDITOR% is >not specified, then OS2ENV.C returns the hardcoded "epm". I can set what I want, crontab refuses to work: [C:\DESKTOP]crontab -e x: No such file or directory crontab: "x" exited with status 1 x is one of my editors but I tried also VIM and e without luck. BTW epm is installed as well but it cannot load it for some reason. Can I debug that somewhat more? cu Adrian -- Adrian Gschwend at OS/2 Netlabs ICQ: 22419590 ktk at netlabs.org ------- The OS/2 OpenSource Project: http://www.netlabs.org **= Email 5 ==========================** Date: Fri, 25 Jan 2002 09:01:45 +0000 From: John Poltorak Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 09:54:37AM +0100, Holger Veit wrote: > On Thu, Jan 24, 2002 at 06:07:59PM +0100, Gerhard Arnecke wrote: > > Concerning header files of BSD one can find this chapter in Toolkit 4.5: > > You probably won't be able to use these files directly in LIBEMU. I > guess the socket support will access the TCP/IP DLLs, but will wrap them > in a similar way as EMX does already now. > > TCP/IP 4.3 brings up the compatibility issue (support for Warp 3 and > pre-FP13 Warp 4) again. I'd like to take advantage of the recent version > of the TCP32 DLLs because these have improved socket support. The old > 4.0 TCP/IP lacks certain functions, and is incompatible, so even systems > with FP14 or later might not yet have a modern TCP/IP stack. I think the 32-bit stack works with Warp 3, and it is a free upgrade AFAICT. > I don't > have a good idea how to deal with that, except by checking the version > and loading the appropriate code on demand. I don't like this because due > to the different functionality, there will be then apps that won't work > on an otherwise perfectly patched Warp 4. If it's perfectly patched, won't that include the 32-bit stack? > Comments? > > Holger > > -- > Please update your tables to my new e-mail address: > holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) > -- John **= Email 6 ==========================** Date: Fri, 25 Jan 2002 09:44:32 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Thu, Jan 24, 2002 at 06:50:22PM +0100, Tobias Huerlimann wrote: > Hi! > > Well, this sounds _very_ promising and I've got some questions on > LIBEMU, just for curiosity. > > On Thu, 24 Jan 2002 15:14:03 +0100, Holger Veit wrote: > > > As it looks now, libemu is the BSD library (Posix/2 already inherited > > quite a lot from this source). You can find the sources > > in a typical FreeBSD or OpenBSD distribution. > > :-) > Which C library do you currently use, Open-, Free-, or NetBSD's one? They do not differ too much for my purposes of testing anything. Currently I have a FreeBSD flavour, but it is not the most recent patch level. I know this, but it has low priority to keep pace with the bleeding edge. > > On porting: I already use the BSD headers for the IX.SYS driver (that's its > > name) where appropriate and possible, but there are some and > > headers which won't be available or will look differently - IX.SYS shares > > some algorithms with the BSD kernel, but is not a BSD kernel itself. > > I wonder which approach you took: > > (1) Did you start modifying a real BSD kernel (i.e. from OpenBSD), > replacing for example all its memory management syscalls using Dos* > calls or > > (2) Did you start rewriting everything from scratch just using some > algorithms? The latter. It does not make too much sense to attempt to extract the low level core of the BSD kernel (like scheduling and memory management and try to map this onto OS/2 structures. This is approximately the microkernel approach which history showed that it is a failure (even though there are still some people who believe in The Hurd mistake). What I do is to establish the environment of the 'struct proc' (see bsd/src/sys/sys/proc.h for reference) side-by-side with OS/2's PTDA management (this already works) and implement the corresponding system calls on top of this. This lives and dies with the access to the internal routine of DosSendSignalException - you can find this quite easily and in a portable way - to be able to issue signals to arbitrary processes and threads. Most relevant system calls which need to deal with struct proc can be found in bsd/src/sys/kern/{kern_prot.c,kern_proc.c,kern_exit.c} and some more I forgot. You can almost copy them 1-to-1 to IX.SYS (note that IX.SYS is with few os2krnl-related 16bit interfaces a rather pure 32 bit C++ driver). Other system calls, like open(), will be handled partly in user mode, by using ordinary wrapped Dos* functions, but need some driver part that does access checking, for instance. The user part will also be responsible to provide Unix compatible virtualized devices, like /dev/mem, /dev/tty, /dev/ttyS0, etc. Long ago, I already sketched here the idea of such virtual devices which translate UNIX ioctls to DosDevIOCtl of real OS/2 devices. Such devices will be only visible in the UnixOS2 environment, but in contrast to access control and symlinks which must be global this is tolerable. > > As the headers are a moving target, compiling the library is a moving > > target as well, and other than getting it statically compile (it won't > > be testable now), there is no real assistance in porting possible. > > Am I right saying that one should be able to take any BSD (or POSIX?) > compliant library and just compile it without any porting (provided > that he has an working 'make' environment)? This is exactly the plan. Configure should recognize a BSD-like system which is almost always supported without problems. Natively, the set of BSD utilities should also build right from the CD ROM. Non-configure-enabled code that does not support a BSD flavour will have the same problems with porting as you would have with FreeBSD/NetBSD/OpenBSD. This usually holds only for software which has to access hardware or kernel features directly, such as access to /dev/kmem. Of course, the whole issue of kernel porting is left out; it won't be possible to port a Unix device driver with this environment without own efforts. The only exception that will need special attention is the gcc compiler and binutils environment which need to be appropriately adapted to OS/2. But there one can profit from the work that is already done in EMX; some day, gcc will be bootstrapped/recompiled with the EMX system, and become an executable that solely links against the LIBEMU DLLs. > > The magic is in the man(2) syscall support > > (this is best comparable to EMX.DLL; here it is the pair IX.SYS and > > UX2KRNL.DLL). > > Will IX.SYS also include XF86SUP.SYS's capabilities and therefore > replace it? Yes, but likely in a more Unixish way, e.g. hidden in some virtual /dev/kmem and /dev/iopl code (or the latter in an iopl() system call). Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 7 ==========================** Date: Fri, 25 Jan 2002 09:54:37 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Thu, Jan 24, 2002 at 06:07:59PM +0100, Gerhard Arnecke wrote: > Concerning header files of BSD one can find this chapter in Toolkit 4.5: You probably won't be able to use these files directly in LIBEMU. I guess the socket support will access the TCP/IP DLLs, but will wrap them in a similar way as EMX does already now. TCP/IP 4.3 brings up the compatibility issue (support for Warp 3 and pre-FP13 Warp 4) again. I'd like to take advantage of the recent version of the TCP32 DLLs because these have improved socket support. The old 4.0 TCP/IP lacks certain functions, and is incompatible, so even systems with FP14 or later might not yet have a modern TCP/IP stack. I don't have a good idea how to deal with that, except by checking the version and loading the appropriate code on demand. I don't like this because due to the different functionality, there will be then apps that won't work on an otherwise perfectly patched Warp 4. Comments? Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 8 ==========================** Date: Fri, 25 Jan 2002 10:22:38 +0000 From: John Poltorak Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 10:59:49AM +0100, Holger Veit wrote: > On Fri, Jan 25, 2002 at 09:01:45AM +0000, John Poltorak wrote: > > On Fri, Jan 25, 2002 at 09:54:37AM +0100, Holger Veit wrote: > > > On Thu, Jan 24, 2002 at 06:07:59PM +0100, Gerhard Arnecke wrote: > > > > Concerning header files of BSD one can find this chapter in Toolkit 4.5: > > > > > > You probably won't be able to use these files directly in LIBEMU. I > > > guess the socket support will access the TCP/IP DLLs, but will wrap them > > > in a similar way as EMX does already now. > > > > > > TCP/IP 4.3 brings up the compatibility issue (support for Warp 3 and > > > pre-FP13 Warp 4) again. I'd like to take advantage of the recent version > > > of the TCP32 DLLs because these have improved socket support. The old > > > 4.0 TCP/IP lacks certain functions, and is incompatible, so even systems > > > with FP14 or later might not yet have a modern TCP/IP stack. > > > > I think the 32-bit stack works with Warp 3, and it is a free upgrade > > AFAICT. > > That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with > Software Choice, MCP/ACP and eCS (so *I* have it), and that it once > in a while leaks through testcase. Do you have a definite URL for > downloading, so that we can fix this together with the the >FP13 > requirement. This will quite simplify the whole procedure. A complete refresh of MPTS is available here:- ftp://service.boulder.ibm.com/ps/products/mpts/fixes/english-us/wr08610_conv/ Here's what README.TXT says:- This FixPak is a refresh of the MPTS LAN Adapter and Protocol Support program and is intended to replace (upgrade) only LICENSED instances of MPTS. Not being a lawyer, I can't be certain about what exactly constitutes a LICENSED instance of MPTS, although I'm not aware of any EOS date for MPTS and it does not appear that the 16-bit stack has being maintained for some time, so it seems as though upgrading to the 32-bit stack is the only option for fixing any bugs expeienced in the 16-bit stack. > > > I don't > > > have a good idea how to deal with that, except by checking the version > > > and loading the appropriate code on demand. I don't like this because due > > > to the different functionality, there will be then apps that won't work > > > on an otherwise perfectly patched Warp 4. > > > > If it's perfectly patched, won't that include the 32-bit stack? > > My impression was that meanwhile certain fixes for drivers (and maybe even > this V4.3 upgrade) are only available if you buy certain support. Are you specifically referring to the v4.3 upgrade or just the 32-bit stack? > I don't want to be so rigid to request something that a user who owns > Warp 4 legally cannot obtain by free download (as I already explained > elsewhere, howewer: if it is available for free, then I request this to > be installed if it is necessary for the job - no mercy with people who > are reluctant to install official patches). > > Holger > > -- > Please update your tables to my new e-mail address: > holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) > -- John **= Email 9 ==========================** Date: Fri, 25 Jan 2002 10:34:52 +0100 (CET) From: "Adrian Gschwend" Subject: BSD stack (was Re: LIBEMU ?) On Fri, 25 Jan 2002 09:54:37 +0100, Holger Veit wrote: >TCP/IP 4.3 brings up the compatibility issue (support for Warp 3 and >pre-FP13 Warp 4) again. I'd like to take advantage of the recent version >of the TCP32 DLLs because these have improved socket support. The old >4.0 TCP/IP lacks certain functions, and is incompatible, so even systems >with FP14 or later might not yet have a modern TCP/IP stack. I don't >have a good idea how to deal with that, except by checking the version >and loading the appropriate code on demand. I don't like this because due >to the different functionality, there will be then apps that won't work >on an otherwise perfectly patched Warp 4. It's not really related to this question but has anyone ever thought about porting a BSD stack (yeah I know, IBM's stack is also BSD) to OS/2? I have several problems with the current 32-bit stack (for example on my system DHCP traps the machine everytime it does the first refresh) so I wonder how much work it would be to port an official BSD stack. Also because I doubt that we will ever get IPv6 from IBM. Not that it would be of much use now but they introduced some very interesting concepts in v6 and I would like to have that on my favorite system as well in a long term. libemu would definitely provide a good base of API's for such a porting I think. For sure it's not done with that, what else do I need to replace the current IP stack? cu Adrian -- Adrian Gschwend at OS/2 Netlabs ICQ: 22419590 ktk at netlabs.org ------- The OS/2 OpenSource Project: http://www.netlabs.org **= Email 10 ==========================** Date: Fri, 25 Jan 2002 10:45:44 +0000 From: John Poltorak Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 11:32:18AM +0100, Holger Veit wrote: > On Fri, Jan 25, 2002 at 10:22:38AM +0000, John Poltorak wrote: > > On Fri, Jan 25, 2002 at 10:59:49AM +0100, Holger Veit wrote: > > > That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with > > > Software Choice, MCP/ACP and eCS (so *I* have it), and that it once > > > in a while leaks through testcase. Do you have a definite URL for > > > downloading, so that we can fix this together with the the >FP13 > > > requirement. This will quite simplify the whole procedure. > > > > A complete refresh of MPTS is available here:- > > > > ftp://service.boulder.ibm.com/ps/products/mpts/fixes/english-us/wr08610_conv/ > > > > Here's what README.TXT says:- > > > > This FixPak is a refresh of the MPTS LAN Adapter and Protocol Support > > program and is intended to replace (upgrade) only LICENSED instances > > of MPTS. > > This is not TCP/IP 4.3. Yes, I know. It's the MPTS stack which was available in TCP/IP 4.1 which was originally only available through Software Choice. > > > My impression was that meanwhile certain fixes for drivers (and maybe even > > > this V4.3 upgrade) are only available if you buy certain support. > > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > stack? > > The complete package which you can download from software choice if you > have a license. My question is related to whether the v4.3 upgrade is an essential requirement or whether the features available in 4.1 are sufficient. I'm not really sure what the real differences are. > Holger > > -- > Please update your tables to my new e-mail address: > holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) > -- John **= Email 11 ==========================** Date: Fri, 25 Jan 2002 10:49:00 +0100 From: Holger Veit Subject: Re: BSD stack (was Re: LIBEMU ?) On Fri, Jan 25, 2002 at 10:34:52AM +0100, Adrian Gschwend wrote: > On Fri, 25 Jan 2002 09:54:37 +0100, Holger Veit wrote: [...] > It's not really related to this question but has anyone ever thought > about porting a BSD stack (yeah I know, IBM's stack is also BSD) to > OS/2? I have several problems with the current 32-bit stack (for > example on my system DHCP traps the machine everytime it does the first > refresh) so I wonder how much work it would be to port an official BSD I think this might rather be a problem with low level network drivers than with the stack logic. > stack. Also because I doubt that we will ever get IPv6 from IBM. Not > that it would be of much use now but they introduced some very > interesting concepts in v6 and I would like to have that on my favorite > system as well in a long term. IPv6 is probably more a "user level" issue than a problem of the current stack. To understand this: IPv6 are normal network packets, so the ethernet adapter will be able to receive and send them (they just have a different ID in the packet header). Likewise to NetBIOS and DecNET support, you'd basically just need a filter that accepts packets with this id, and pushes it up to the higher ISO layers. The IPv6 socket support is mainly interpretation of addresses and header information. Such support is not in the current TCP/IP 32 libraries yet, but once you be able to attach to the packet filter device, you can extend it. You don't need to replace the whole stack. > libemu would definitely provide a good base of API's for such a porting > I think. For sure it's not done with that, what else do I need to > replace the current IP stack? -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 12 ==========================** Date: Fri, 25 Jan 2002 10:52:08 +0000 From: John Poltorak Subject: Re: BSD stack (was Re: LIBEMU ?) On Fri, Jan 25, 2002 at 10:34:52AM +0100, Adrian Gschwend wrote: > It's not really related to this question but has anyone ever thought > about porting a BSD stack (yeah I know, IBM's stack is also BSD) to > OS/2? I have several problems with the current 32-bit stack (for > example on my system DHCP traps the machine everytime it does the first > refresh) You should obtain a fix or at least report the problem. DHCP should not trap. If it does, then it's either misconfigured or you have found a bug. If it's a bug, you would expect that others have also found it and possibly reported it. > cu > > Adrian > > > -- > Adrian Gschwend > at OS/2 Netlabs > > ICQ: 22419590 > ktk at netlabs.org > ------- > The OS/2 OpenSource Project: > http://www.netlabs.org > -- John **= Email 13 ==========================** Date: Fri, 25 Jan 2002 10:59:49 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 09:01:45AM +0000, John Poltorak wrote: > On Fri, Jan 25, 2002 at 09:54:37AM +0100, Holger Veit wrote: > > On Thu, Jan 24, 2002 at 06:07:59PM +0100, Gerhard Arnecke wrote: > > > Concerning header files of BSD one can find this chapter in Toolkit 4.5: > > > > You probably won't be able to use these files directly in LIBEMU. I > > guess the socket support will access the TCP/IP DLLs, but will wrap them > > in a similar way as EMX does already now. > > > > TCP/IP 4.3 brings up the compatibility issue (support for Warp 3 and > > pre-FP13 Warp 4) again. I'd like to take advantage of the recent version > > of the TCP32 DLLs because these have improved socket support. The old > > 4.0 TCP/IP lacks certain functions, and is incompatible, so even systems > > with FP14 or later might not yet have a modern TCP/IP stack. > > I think the 32-bit stack works with Warp 3, and it is a free upgrade > AFAICT. That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with Software Choice, MCP/ACP and eCS (so *I* have it), and that it once in a while leaks through testcase. Do you have a definite URL for downloading, so that we can fix this together with the the >FP13 requirement. This will quite simplify the whole procedure. > > I don't > > have a good idea how to deal with that, except by checking the version > > and loading the appropriate code on demand. I don't like this because due > > to the different functionality, there will be then apps that won't work > > on an otherwise perfectly patched Warp 4. > > If it's perfectly patched, won't that include the 32-bit stack? My impression was that meanwhile certain fixes for drivers (and maybe even this V4.3 upgrade) are only available if you buy certain support. I don't want to be so rigid to request something that a user who owns Warp 4 legally cannot obtain by free download (as I already explained elsewhere, howewer: if it is available for free, then I request this to be installed if it is necessary for the job - no mercy with people who are reluctant to install official patches). Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 14 ==========================** Date: Fri, 25 Jan 2002 11:21:01 +0100 (CET) From: "Adrian Gschwend" Subject: Re: LIBEMU ? On Fri, 25 Jan 2002 10:59:49 +0100, Holger Veit wrote: >That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with >Software Choice, MCP/ACP and eCS (so *I* have it), and that it once >in a while leaks through testcase. Do you have a definite URL for >downloading, so that we can fix this together with the the >FP13 >requirement. This will quite simplify the whole procedure. It isn't free. It leaked from IBM once as you write but you cannot get the package on a SW-Choice page without a subscription. For sure there are other channels where you can get it but they are not legal so not a real option. cu Adrian -- Adrian Gschwend at OS/2 Netlabs ICQ: 22419590 ktk at netlabs.org ------- The OS/2 OpenSource Project: http://www.netlabs.org **= Email 15 ==========================** Date: Fri, 25 Jan 2002 11:32:18 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 10:22:38AM +0000, John Poltorak wrote: > On Fri, Jan 25, 2002 at 10:59:49AM +0100, Holger Veit wrote: > > That 'AFAICT' is my concern. I know that you'll get TCP/IP V4.3 with > > Software Choice, MCP/ACP and eCS (so *I* have it), and that it once > > in a while leaks through testcase. Do you have a definite URL for > > downloading, so that we can fix this together with the the >FP13 > > requirement. This will quite simplify the whole procedure. > > A complete refresh of MPTS is available here:- > > ftp://service.boulder.ibm.com/ps/products/mpts/fixes/english-us/wr08610_conv/ > > Here's what README.TXT says:- > > This FixPak is a refresh of the MPTS LAN Adapter and Protocol Support > program and is intended to replace (upgrade) only LICENSED instances > of MPTS. This is not TCP/IP 4.3. > > My impression was that meanwhile certain fixes for drivers (and maybe even > > this V4.3 upgrade) are only available if you buy certain support. > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > stack? The complete package which you can download from software choice if you have a license. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 16 ==========================** Date: Fri, 25 Jan 2002 11:50:31 +0100 (CET) From: "Christian Hennecke" Subject: Re: netCDF On Thu, 24 Jan 2002 09:52:23 +0000, John Poltorak wrote: >It's great to come across OS/2 ports like this even though it isn't >something I have a need for, unless someone familiar with the package >could suggest what I could use it for... NetCDF is used by the Generic Mapping Tools to process data. A great set of utilities for visualizing spatial data BTW. Christian Hennecke **= Email 17 ==========================** Date: Fri, 25 Jan 2002 12:05:01 +0100 (CET) From: "Adrian Gschwend" Subject: Re: BSD stack (was Re: LIBEMU ?) On Fri, 25 Jan 2002 10:52:08 +0000, John Poltorak wrote: (it becomes OT, sorry :-) >You should obtain a fix or at least report the problem. DHCP should not >trap. If it does, then it's either misconfigured or you have found a bug. >If it's a bug, you would expect that others have also found it and >possibly reported it. well I'm not the only one with this bug, I got it for month now and I also have other bugs with the IP stack. For example I often change the network with my laptop without rebooting (thanks to OS/2 :-). That means other IP address and other default GW's so I change that by some REXX scripts. From time to time changing the default GW is enough to get a trap in AFINET driver. That sucks bigtime. Knut had similar problems as well and he thinks it might be an overflow somewhere but he couldn't locate it exactly so far. IBM knows the problem but cannot reproduce it as it looks like I guess it's a combination of driver and IP stack. cu Adrian -- Adrian Gschwend at OS/2 Netlabs ICQ: 22419590 ktk at netlabs.org ------- The OS/2 OpenSource Project: http://www.netlabs.org **= Email 18 ==========================** Date: Fri, 25 Jan 2002 12:05:21 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 10:45:44AM +0000, John Poltorak wrote: > > This is not TCP/IP 4.3. > > Yes, I know. It's the MPTS stack which was available in TCP/IP 4.1 which > was originally only available through Software Choice. > > > > > > My impression was that meanwhile certain fixes for drivers (and maybe even > > > > this V4.3 upgrade) are only available if you buy certain support. > > > > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > > stack? > > > > The complete package which you can download from software choice if you > > have a license. > > My question is related to whether the v4.3 upgrade is an essential > requirement or whether the features available in 4.1 are sufficient. I'm > not really sure what the real differences are. I was under the impression that the 4.x (x > 0) version contains DLLs which export entries like socketpair() and which have improved AF_UNIX support. This is not in so32dll.dll, but in a different DLL. I just don't know whether x=1 or x=3 is required. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 19 ==========================** Date: Fri, 25 Jan 2002 12:44:55 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 06:31:47AM -0500, Jack Troughton wrote: > Holger Veit wrote: > > I was under the impression that the 4.x (x > 0) version contains DLLs > > which export entries like socketpair() and which have improved AF_UNIX > > support. This is not in so32dll.dll, but in a different DLL. I just > > don't know whether x=1 or x=3 is required. > > Hmmmm.... > > [C:\mptn\dll]exeinfo -x tcpip32.dll | grep -i pair > > ExeInfo, New Exe File Format Info, v2.0f > (C) Copyright Ngb Technologies, 1991-1993, All rights reserved. > > 254 SOCKETPAIR > > I think it's in tcpip32.dll :) Thanks, I have seen it, but have had no access to an OS/2 machine here at work to check it. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 20 ==========================** Date: Fri, 25 Jan 2002 12:47:56 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Fri, Jan 25, 2002 at 06:25:56AM -0500, Jack Troughton wrote: [...] > > > Are you specifically referring to the v4.3 upgrade or just the 32-bit > > > stack? > > > > The complete package which you can download from software choice if you > > have a license. > > Socket support is in the MPTS refresh that John mentions above. Most of > the changes between 4.1 and 4.3 is not in the stack, but in the > applications included with the stack; for example, in 4.3 one gets dhcp > and ddns servers. > > This is evidenced by the fact that whether or not one has 4.1 or 4.3, > you have to apply MPTN fix IC27649 in order to get rid of the "trap > after reconnecting a ppp connection" problem. > > You should be able to tell people to get 8620 and then ic27649; it is > legal and should give you the services you need. It's the applications > that IBM is protecting, not the stack itself. Thanks for the clarification. This sounds good. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 21 ==========================** Date: Fri, 25 Jan 2002 12:50:31 +0000 From: John Poltorak Subject: Autoconf 2.52g is released **= Email 22 ==========================** Date: Fri, 25 Jan 2002 13:11:16 +0100 (MEZ) From: "Gerhard Arnecke" Subject: Re: LIBEMU ? Oliver Rick has listed the important fixes for TCP/IP v4.3x: http://www.warpupdates.mynetcologne.de/english/tcpip43.html The Text of IC27649 : ------------------- A trap occurs in afinet.sys while bringing down a ppp connection. Installation: ------------ Make a backup of the sys files from \mptn\protocol directory and dll files from \mptn\dll directory. Restart the machine in minimal bootup mode. (i.e. when the machine is restarting and when a small white box appears on the left top corner of the screen, hit F1 and then F2). Copy the new sys files to \mptn\protocol directory and sys files to \mptn\dll directory Restart the machine. Files ----- 15/11/00 4,38p 19652 0 afinet.sym 15/11/00 4,38p 309248 61 afinet.sys 15/11/00 4,38p 19668 0 afinetk.sym 15/11/00 4,38p 308224 61 afinetk.sys 15/11/00 4,38p 15892 0 aflean.sym 15/11/00 4,38p 209920 61 aflean.sys 15/11/00 4,38p 3652 0 afos2.sym 15/11/00 4,38p 13312 61 afos2.sys 15/11/00 4,38p 1566 0 readme 15/11/00 4,38p 64673 0 route.c 15/11/00 4,38p 14397 0 so32dll.dll 15/11/00 4,38p 10964 0 sockets.sym 15/11/00 4,38p 116224 61 sockets.sys 15/11/00 4,38p 11844 0 socketsk.sym 15/11/00 4,38p 124416 61 socketsk.sys 15/11/00 4,38p 19293 0 tcp32dll.dll 15/11/00 4,38p 98332 61 tcpip32.dll 15/11/00 4,38p 14344 0 tcpipdll.dll With MPTS updates 8610,8620 and 8621 one gets the unified level 32 bit. Oliver Rick presents this in a specifuc feature: http://www.warpupdates.mynetcologne.de/english/net_mpts.html **= Email 23 ==========================** Date: Fri, 25 Jan 2002 14:20:28 +0300 (MSK) From: "Andrew Belov" Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error? On Fri, 25 Jan 2002 10:57:57 +0300 (MSK), Adrian Gschwend wrote: > I can set what I want, crontab refuses to work: > > [C:\DESKTOP]crontab -e > x: No such file or directory > crontab: "x" exited with status 1 Just forgot to mention, crontab requires a KSH/BASH-alike shell, the one that will treat "-c" in the same way as CMD.EXE treats "/c". The shell is specified to VXCron with %_PATH_BSHELL% environment variable and defaults to "sh". Since there are many tools using "sh", I just put the Korn shell 5.2.7 under SH.EXE, but you may also set _PATH_BSHELL to KSH.EXE or BASH.EXE if you like. That is of litte relevance unless your editor deploys some shell scripts too. The error message may be ambiguous but that's how do they run execlp(). **= Email 24 ==========================** Date: Fri, 25 Jan 2002 20:41:35 -0700 (MST) From: "Tim Erickson" Subject: Error with Perl 561 I have been trying to get perl running on my system and testperl.cmd returns results similar to the following when run from the commandline: [I:\perl\bin\5.6.1]perl___.exe Process terminated by SIGILL SYS1808: The process has stopped. The software diagnostic code (exception code) is 001C. [I:\perl\bin\5.6.1]perl Process terminated by SIGILL SYS1808: The process has stopped. The software diagnostic code (exception code) is 001C. and when perlPM is run I get a sys3176 in perld12e.dll. Any ideas why? Tim **= Email 25 ==========================** Date: Fri, 25 Jan 2002 23:56:41 -0500 From: Henry Sobotka Subject: Re: Error with Perl 561 Yes, I screwed up by confusing gcc's -mcpu and -march flags. Replacements are in http://www.elzahir.org/perl/. They're just the binaries, so install by overwriting the original installation. h~