From: UnixOS2 Archive To: "UnixOS2 Archive" Date: Wed, 23 Jan 2002 04:09:45 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 113 ************************************************** Tuesday 22 January 2002 Number 113 ************************************************** Subjects for today 1 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : John Poltorak 2 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : Andrew Belov" 3 Re: Real Audio for Unix : Jack Troughton 4 Re: jpeg6 & libpng : Andrew MacIntyre 5 Re: New Python - MailMan : Andrew MacIntyre 6 Re: Real Audio for Unix : csaba.raduly at sophos.com 7 Re: Install Compatibility : John Poltorak 8 Re: Real Audio for Unix : John Poltorak 9 Re: Reading AIX diskette images : John Poltorak 10 Make 3.79.1 : John Poltorak 11 Re: sox 12.17.3 : John Poltorak 12 sox 12.17.3 : Michel SUCH" 13 LIBEMU ? : John Poltorak 14 Re: LIBEMU ? : Stefan Neis 15 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : John Poltorak 16 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : John Poltorak 17 Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) : Adrian Gschwend" 18 Re: LIBEMU ? : Holger Veit **= Email 1 ==========================** Date: Wed, 23 Jan 2002 00:19:15 +0000 From: John Poltorak Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Wed, Jan 23, 2002 at 01:27:18AM +0300, Andrew Belov wrote: > On Tue, 22 Jan 2002 14:58:11 -0600, email at eracc.hypermart.net wrote: > > >What is the chance that we can get an 'at' that works with vxcron or > >is one supplied? > > VXCron is not bundled with an at/atd suite, or at least it's not bundled this way in the > BlackCat Linux v 6.2 package. There is at-3.1.7 on the BlackCat CDs, with a variety of > patches. But I don't know if it is supposed to interoperate with VXCron... It would be nice if you could port this one:- ftp://ftp.mirror.ac.uk/sites/ftp.slackware.com/pub/slackware/slackware/source/a/bin/at_3.1.8.orig.tar.gz > at least I'm quite > happy with the IBM genuine AT.EXE from LAN Server. ;-) This seems very much limited compared with ATD... -- John **= Email 2 ==========================** Date: Wed, 23 Jan 2002 01:27:18 +0300 (MSK) From: "Andrew Belov" Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Tue, 22 Jan 2002 14:58:11 -0600, email at eracc.hypermart.net wrote: >What is the chance that we can get an 'at' that works with vxcron or >is one supplied? VXCron is not bundled with an at/atd suite, or at least it's not bundled this way in the BlackCat Linux v 6.2 package. There is at-3.1.7 on the BlackCat CDs, with a variety of patches. But I don't know if it is supposed to interoperate with VXCron... at least I'm quite happy with the IBM genuine AT.EXE from LAN Server. ;-) **= Email 3 ==========================** Date: Wed, 23 Jan 2002 07:49:47 -0500 From: Jack Troughton Subject: Re: Real Audio for Unix John Poltorak wrote: > > On Wed, Jan 23, 2002 at 10:09:25AM +0000, csaba.raduly at sophos.com wrote: > > > > On 22/01/2002 18:57:24 owner-os2-unix wrote: > > > > >How do Linux users handle Real Audio files? > > > > > > > Badly. They bitch and moan about proprietary file formats :-) > > Or they run Wine, VMware or suchlike. > > Is any Open Source Software available for streaming audio and video likely > to become established as standard formats for such media? Depends on what you mean by standard format... I don't think it's likely that most North American and/or Western European media outlets are likely to use them in the immediate future, because they like the idea of control that the proprietary solutions (seem to) provide. > I'm aware of Ogg Vorbis for audio and Ogg Tarkin for video (see > http://www.xiph.org/ogg) but have no idea if they are ever likely to > become generally available on Web sites. They ought to be, before > Microsoft completely dictates what anyone can do on the Internet. I think you'll find that fake people such as large corporations and the like will use real, and that real people will use fake (errr... I mean open source stuff;). Regards, Jack **= Email 4 ==========================** Date: Wed, 23 Jan 2002 07:59:37 +1100 (EDT) From: Andrew MacIntyre Subject: Re: jpeg6 & libpng On Tue, 22 Jan 2002, Stefan Neis wrote: > On Mon, 21 Jan 2002, John Poltorak wrote: > > > > Although the libpng package has DLL support, I'm taking the approach of > > > only preferring static libraries for the Python loadable modules (which > > > are themselves DLLs) to minimise the risk of hard to track down problems > > > with my Python binary packages. > > > HCChu... He used different FLAGS to you and the binaries were much smaller > > not sure if that is necessarily a consequence of the flags. > > Also, this sounds like those DLLs contain some Python specific extensions. The Python loadable modules are Python specific, even if they are OS/2 DLLs. They can't really be used by any other code without dragging in and intitialising the Python interpreter DLL. As far as my versions of libjpeg and libpng are concerned, there is nothing Python specific in them. However my choice of compiler options is geared towards integration with my Python binary distributions. > It might be a good idea to have those included in the mainstream DLL > instead of either writing a wrapper DLL calling the plain C DLL or having > a second version of DLL containing a static version of the C DLL, but it's > actually not a totally trivial decision: > - Is this functionality sufficiently stable (Python itself still seems to > be evolving rapidly) to have it in a system level DLL? > - Is the overhead for Python specific extension small enough to want to > impose it on _all_ UnixOS/2 users, even those not using Python? > - This seems to leave the idea of just blindly mirroring Unix distros > behind. However, it might be a good idea to not blindly follow other > people's (possibly bad) ideas. While having some of this linked static into Python's loadable modules bloats Python a bit, it saves users of my binary distribution lots of headaches in dealing with the plethora of DLLs (like zlib) floating around. My next binary release will have zlib linked static into Python's zlib loadable module to escape the current dependency. UnixOS2 is free to setup a Python distribution that uses the DLLs you are distributing. I don't have a problem with that, and nothing (that I'm aware of) in the Python OS/2 build infrastructure should get in the way of that. But I don't consider it appropriate for my binary distribution, which I am trying to make as "batteries included" as possible for the widest range of OS/2 users. > Oh yes, and last, but most important: > - Does it export all the symbols that are exported by standard DLL/.so > files or does it export only those symbols needed by Python. I don't recall changing the .DEF file in the libpng source distribution for my port, so this shouldn't be an issue. Regards, Andrew. -- Andrew I MacIntyre "These thoughts are mine alone..." E-mail: andymac at bullseye.apana.org.au | Snail: PO Box 370 andymac at pcug.org.au | Belconnen ACT 2616 Web: http://www.andymac.org/ | Australia **= Email 5 ==========================** Date: Wed, 23 Jan 2002 08:04:36 +1100 (EDT) From: Andrew MacIntyre Subject: Re: New Python - MailMan On Tue, 22 Jan 2002, IanM wrote: > I've been following the manual install instructions, I could > spend the time rewritting the make install, only when I've looked > at it, after the gasp, I've always continued reading the manual > install bits. Yes, some of the Python in the configure script needs modification to do the "configure" part of the build correctly. If I complete my port, I'll submit those changes to the Mailman maintainer. As I think I've said before, I'm focussing on getting the EMX port into Python's CVS at the moment. Then I might get back to Mailman and Zope. -- Andrew I MacIntyre "These thoughts are mine alone..." E-mail: andymac at bullseye.apana.org.au | Snail: PO Box 370 andymac at pcug.org.au | Belconnen ACT 2616 Web: http://www.andymac.org/ | Australia **= Email 6 ==========================** Date: Wed, 23 Jan 2002 10:09:25 +0000 From: csaba.raduly at sophos.com Subject: Re: Real Audio for Unix On 22/01/2002 18:57:24 owner-os2-unix wrote: >How do Linux users handle Real Audio files? > Badly. They bitch and moan about proprietary file formats :-) Or they run Wine, VMware or suchlike. -- Csaba Ráduly, Software Engineer Sophos Anti-Virus email: csaba.raduly at sophos.com http://www.sophos.com US Support: +1 888 SOPHOS 9 UK Support: +44 1235 559933 **= Email 7 ==========================** Date: Wed, 23 Jan 2002 10:32:40 +0000 From: John Poltorak Subject: Re: Install Compatibility On Tue, Jan 22, 2002 at 09:16:22PM -0500, jimmy wrote: > Hello > I first considered sending this to the OS2-Linux group but it seems less active > and serious than this group and I hope it applies here. The os2-linux group seems to be more focused on migrating from OS/2 to Linux, whereas the principal aim here is to provide as much Linux or Unix functionality as possible on OS/2 without any intention of migrating, even though any expertise gained may well be useful at some future point in dealing with Linux. A lot of people here would argue that OS/2 is far more mature than Linux so what is the point in migrating especially if any Open Source Software can be made to run on OS/2. > Recently I became aware > of a couple of Slackware projects ( now available for debian, too ) that may help > us out involving installation/uninstallation. I have always relied on compiling > from source so had no interest in installers but now "Checkinstall" and it's > associated utilities ( like "installwatch") has been assembled in a manner that it > is run just before "make install", resulting in a log and only a small step to a > "slackpak" for "pkgtool", so that test installations, clean upgrades etc are > possible without having to turn over the work to a > "do-everything-behind-your-back" binary.... clean and neat. Current version is > 1.5.1. Considering we seem to be using slack as something of a model I thought it > worthwhile to mention these projects in hopes that someone stronger than I can put > them to use for UnixOS2. Slackware is being used as a reference platform from the POV of building up a distro and installer. It does save a lot of discussion on directory structures and package management etc. However the aim is to have a distro which is FHS compliant and I'm not sure that Slackware is, so that at some point we may need to review the current directory structure currently adopted. > If this is already commonly known, please forgive the > redundancy. If not and you think they may be of some use,"checkinstall", > "installwatch" ( usually included ), "spasm", etc. are available from > freshmeat.net. I've just begun working with them so have little else to offer at > this point but interest, and direction. Hopefully that will soon accelerate and > grow. It's useful having input like this. We currently do have a UnixOS/2 installer based on PKGTOOL but I suspect it hasn't been used very much as the distro is still getting built up. I'm not familiar with the packages mentioned but if anyone thinks they would make life easier in getting UnixOS/2 up and running, then they are worth considering. > Jimmy > > > -- John **= Email 8 ==========================** Date: Wed, 23 Jan 2002 10:47:16 +0000 From: John Poltorak Subject: Re: Real Audio for Unix On Wed, Jan 23, 2002 at 10:09:25AM +0000, csaba.raduly at sophos.com wrote: > > On 22/01/2002 18:57:24 owner-os2-unix wrote: > > >How do Linux users handle Real Audio files? > > > > Badly. They bitch and moan about proprietary file formats :-) > Or they run Wine, VMware or suchlike. Is any Open Source Software available for streaming audio and video likely to become established as standard formats for such media? I'm aware of Ogg Vorbis for audio and Ogg Tarkin for video (see http://www.xiph.org/ogg) but have no idea if they are ever likely to become generally available on Web sites. They ought to be, before Microsoft completely dictates what anyone can do on the Internet. > > -- > Csaba Ráduly, Software Engineer Sophos Anti-Virus > email: csaba.raduly at sophos.com http://www.sophos.com > US Support: +1 888 SOPHOS 9 UK Support: +44 1235 559933 > -- John **= Email 9 ==========================** Date: Wed, 23 Jan 2002 11:20:35 +0000 From: John Poltorak Subject: Re: Reading AIX diskette images On Tue, Jan 22, 2002 at 01:43:24PM +0100, Gerhard Arnecke wrote: > Download the tool DSK4xxx (v4.22 for OS/2 from the title) from this site: > > http://dvalot.free.fr/emtcopy.htm > > Furthermore click on other other site and downlaod the documentation. I have looked around and can manage to extract files from a diskette image to an SVDISK, but I can't find any way of seeing the filesystem. I guess what I'd really like is an image to ZIP converter... Wonder if it's possible to use DD to read the diskette... but even then the output would still not allow me to view it under OS/2. Maybe tar could be used... > GA > > > -- John **= Email 10 ==========================** Date: Wed, 23 Jan 2002 16:35:11 +0000 From: John Poltorak Subject: Make 3.79.1 There was a test version of Make 3.79.1 made early last year. Does anyone know where it was uploaded? ISTR netlabs or unixos2.com but can't locate it now. -- John **= Email 11 ==========================** Date: Wed, 23 Jan 2002 19:21:16 +0000 From: John Poltorak Subject: Re: sox 12.17.3 On Wed, Jan 23, 2002 at 07:54:51PM +0100, Michel SUCH wrote: > Hi all, > > I can successfully compile this sox version, but it does not seem to be > able to correctly parse its parameters. > > For example, with version 12.17.1, if you want to convert a stereo wave > file to mono, you should enter: > sox stereo.wav -c 1 mono.wav > > This does not work with nor 12.17.2 neither with 12.17.3 which seems to > think that -c is the name of the output file. > Has anyone tried this? Is it a bug or a problem with porting? I just tried it and got:- sox.exe: Unknown output file format for '-c': Filetype was not specified It may be worth reporting on the SoX mailing list:- http://lists.sourceforge.net/lists/listinfo/sox-users > ---------------------------- > Michel SUCH TEAM OS/2 FRANCE > ICQ # 51654489 > -- John **= Email 12 ==========================** Date: Wed, 23 Jan 2002 19:54:51 +0100 (CET) From: "Michel SUCH" Subject: sox 12.17.3 Hi all, I can successfully compile this sox version, but it does not seem to be able to correctly parse its parameters. For example, with version 12.17.1, if you want to convert a stereo wave file to mono, you should enter: sox stereo.wav -c 1 mono.wav This does not work with nor 12.17.2 neither with 12.17.3 which seems to think that -c is the name of the output file. Has anyone tried this? Is it a bug or a problem with porting? ---------------------------- Michel SUCH TEAM OS/2 FRANCE ICQ # 51654489 **= Email 13 ==========================** Date: Wed, 23 Jan 2002 21:10:32 +0000 From: John Poltorak Subject: LIBEMU ? Here's part of a conversation from USENET:- On Wed, Jan 23, 2002 at 12:58:04PM -0800, Ilya Zakharevich wrote: > [A complimentary Cc of this posting was sent to > John Poltorak > IIUC, the idea of LIBEMU is > > If you want grep 3.0.1 (as opposed to 3.0.0) you need to install a > device driver which will interfare with *almost all* functions of > your system, > Basically, such a device driver *by itself* is a very good thing. > Having a CRT library which *can* interface with such a driver is a > very good thing. > > *Requiring* such a device driver for *any Unix-like* application is a > lunacy. > Ilya None of this sounds like the sort of thing I envisage. My understanding is that LIBEMU will provide an alternative to EMXRT. Where do this notion of device driver come from? -- John **= Email 14 ==========================** Date: Wed, 23 Jan 2002 22:38:17 +0100 (CET) From: Stefan Neis Subject: Re: LIBEMU ? On Wed, 23 Jan 2002, John Poltorak wrote: > > > IIUC, the idea of LIBEMU is > > > > If you want grep 3.0.1 (as opposed to 3.0.0) you need to install a > > device driver which will interfare with *almost all* functions of > > your system, (snipp) > > *Requiring* such a device driver for *any Unix-like* application is a > > lunacy. > None of this sounds like the sort of thing I envisage. My understanding is > that LIBEMU will provide an alternative to EMXRT. > > Where do this notion of device driver come from? I guess that's an allusion to the idea of implicitly "rewriting" all the path names and environment variables (e.g. make /etc/bla access %UNIXROOT%/etc/bla and get rid of drive letters). Also some specific Unix functions will probably require low level (i.e. device driver access) to the system and as LIBEMU can't decide whether a given application will require any of those advanced functions or not, it won't work without the device driver installed... Holger, did I get it approximately right? Regards, Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 15 ==========================** Date: Wed, 23 Jan 2002 23:12:20 +0000 From: John Poltorak Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Thu, Jan 24, 2002 at 12:01:46AM +0100, Adrian Gschwend wrote: > On Wed, 23 Jan 2002 23:48:09 +0100 (CET), Adrian Gschwend wrote: > > >I want to start several jobs, like some cmd-scripts but also rm -r on > >the temp directory. > >Should I now write a cmd-script which executes rm -r or should I simply > >enter the command itself in the file? Might be a stupid question but I > >just wonder. > > hmm just started crontab and it cannot find epm, I think I don't have > that installed IIRC. Does it rely on that or can I change that? Set the EDITOR variable to your chosen editor. > 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 16 ==========================** Date: Wed, 23 Jan 2002 23:33:27 +0000 From: John Poltorak Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Thu, Jan 24, 2002 at 12:19:10AM +0100, Adrian Gschwend wrote: > On Wed, 23 Jan 2002 23:12:20 +0000, John Poltorak wrote: > > >Set the EDITOR variable to your chosen editor. > > yeah noted that later :-) see other post > > but: > > [C:\mptn\etc\cron]crontab -e > epm: No such file or directory > crontab: "epm" exited with status 1 > > that happens with every editor I set. Does it work on your machine? Yes. I guess it has something to do with the environment. crontab appears to take a debug option but I can't work out how to invoke it:- [N:\eval\cron-vx\bin]crontab -x -u root -l unrecognized debug flag <-u> <-u> crontab: usage error: bad debug option > 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 17 ==========================** Date: Wed, 23 Jan 2002 23:48:09 +0100 (CET) From: "Adrian Gschwend" Subject: Re: vxcron (was Re: Cron/2 1.41 DOS Session error?) On Tue, 22 Jan 2002 14:58:11 -0600, email at eracc.hypermart.net wrote: >>There is also a minor fix to tolerate the presence of subdirectories in >>%CRONDDIR%, allowing to have the crontabs under CVS version control. >>Read the relevant manpage ("man 8 cron") for more details. > >Oh goody! I'm going to go get and try this one. Thanks Andrew! got it as well, looks very good to me. I'm quite new to cron so I have a question about the usage: I want to start several jobs, like some cmd-scripts but also rm -r on the temp directory. Should I now write a cmd-script which executes rm -r or should I simply enter the command itself in the file? Might be a stupid question but I just wonder. >What is the chance that we can get an 'at' that works with vxcron or >is one supplied? I guess I'll know the answer to the latter in a few. >Thanks again! I know it's always easy to ask for ports instead of doing it myself but I vote for that one as well :-)) But the past weeks I invested a lot of time into the netlabs.org redesign (backend mostly) so no time to read the porting guide yet :) 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: Wed, 23 Jan 2002 23:55:41 +0100 From: Holger Veit Subject: Re: LIBEMU ? On Wed, Jan 23, 2002 at 10:38:17PM +0100, Stefan Neis wrote: > On Wed, 23 Jan 2002, John Poltorak wrote: > > > > > IIUC, the idea of LIBEMU is > > > > > > If you want grep 3.0.1 (as opposed to 3.0.0) you need to install a > > > device driver which will interfare with *almost all* functions of > > > your system, > (snipp) > > > *Requiring* such a device driver for *any Unix-like* application is a > > > lunacy. > > > None of this sounds like the sort of thing I envisage. My understanding is > > that LIBEMU will provide an alternative to EMXRT. > > > > Where do this notion of device driver come from? > > I guess that's an allusion to the idea of implicitly "rewriting" all > the path names and environment variables (e.g. make /etc/bla access > %UNIXROOT%/etc/bla and get rid of drive letters). Also some specific > Unix functions will probably require low level (i.e. device driver > access) to the system and as LIBEMU can't decide whether a given > application will require any of those advanced functions or not, > it won't work without the device driver installed... > > Holger, did I get it approximately right? Close, but not full match. In that usenet thread, I already gave a glance of what this driver can and will accomplish. There is always the talk about the UNIXROOT redirection - infact, this is a pretty simple and unimportant task, and it could even be managed by user level means, though at the price of having a local effect only: e.g. UnixOS/2 apps would understand symbolic links, but other apps won't, leading to the not nice result that you'd edit two different files, depending on the editor (UnixOS/2 or non-UNIXOS2 one) chosen. It is also not an issue as with XF86SUP.SYS which provides certain I/O or memory access which wouldn't be possible without the help of a driver. It is much more involved: the driver will provide the Unix man(2) API, where functions are missing in the os2krnl, and will modify behaviour of existing Dos* calls where Unix issues get involved. Some examples: Unix has users and groups and enforces access to files based on process ownership. You could emulate this with a tricky mechanism in EMX as well (currently, EMX doesn't have any of this), if you'd maintain a global memory area and a manager process that is queried for any file access (file access rights could be kept in EAs). This would give users the impression that there is Unix access control in UnixOS2. Unfortunately, even a simple DEL command, issued by the CMD interpreter will destroy this impression - outside the UnixOS/2 shell, you still remain superuser. You can forget about such kind of "security" entirely. You'll get real security only if it is globally handled, and if all code running in the system is controlled. This requires a special driver, an ISS (installable security system) BASEDEV. IX.SYS is such an ISS. You can look at a simple sample ISS in the MWDD32 sources, but what this code will allow (to intercept about 30 routines, like open, read, exec, settime, etc.) is not sufficient. The real undocumented component (which doesn't mean it doesn't work) are two unsuspicious callbacks which allow to control almost any Dos API call. This is badly needed - the official SES API was apparently written explicitly for things like virus scanners (software that just monitors certain events) - there is for instance a callback that allows to verify whether some process may be run, and whether it is started actually, but the inverse call that allows to check whether that process has terminated is missing. The latter is not only interesting to implement the accounting API acct(2), but also to find out what processes are actually running at all currently, without continuously polling DosQuerySysState (which is a slow call, which makes it unsuitable by use by a user mode monitor process). The information about all processes is important to implement the family Unix process group and session calls (setpgrp, setpgid, setsid) and to get correct Unix signal handling. One of the most annoying problems in XFree86 is that terminating the X server will usually lead to remaining X clients which where started detached, and thus get inherited by the PMSHELL session. PMSHELL is OS/2's SESMGR, session manager; this is an ugly hack which is contained in DOSCALL1.DLL. If you want to get rid of such detached processes, you'll have to maintain a list of them somewhere and explicitly kill them (with the "death.exe" kill function). This is best done from the kernel level itself - any outside component like the assumed global manager will conflict with security issues. You close a single hole somewhere, but open a different hole elsewhere. Symbolic links are another global domain you better leave to a kernel component, for the editor argument mentioned above. In the usenet thread, I already described what TVFS (which is already somehow a kernel component - an IFS) actually works; this relies on a user mode process to do the whole work. This is not only slow and a systemwide bottleneck; it also contains a builtin limit of the maximum number of files maintainable (the per process file table is limited to a 16 bit 64K segment which is large, but not unlimited). Finally, there is the fork()/exec() complex which is a complicated and tricky mechanism in EMX. It is infact miraculous what EM built there; and it is quite hard to impossible to improve this with user-level methods. At kernel level, it is possible to do much more. I just point to the Odin driver which manages to integrate the foreign PE/COFF executable format into the system transparently. Hope this clears up the obscure driver thing a bit. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection)