From: UnixOS2 Archive To: "UnixOS2 Archive" Date: Fri, 26 Jul 2002 04:33:37 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 284 ************************************************** Thursday 25 July 2002 Number 284 ************************************************** Subjects for today 1 Re: UnixOS/2 bootstrap : Maynard" 2 Re: wxWindows : Neil Waldhauer 3 RE: wxWindows : Dave Webster 4 bootstrap.lst : Maynard" 5 Re: Specifying YACC as preference over BISON : Thomas E. Dickey" 6 RE: wxWindows : Dave Webster 7 RE: Multiplatform LS : Dave Webster 8 GLIBC : John Poltorak 9 Re: ln.cmd : Stefan Neis 10 Re: wxWindows : Jack Troughton 11 baseline perl harness results : Maynard" 12 GCC =?iso-8859-1?q?3.2=3F?= : Mika Kristian Laitio 13 Re: GLIBC : Stefan Neis 14 Re: GCC =?iso-8859-1?q?3.2=3F?= : Stefan Neis 15 Re: UnixOS/2 bootstrap : Dave Saville" 16 Multiplatform LS : Lord Spigol" 17 RE: wxWindows : Stefan Neis 18 Baseline EMX [emx_inst.cmd] : Maynard" 19 Re: bootstrap.lst : Maynard" 20 Re: UnixOS/2 bootstrap : John Poltorak 21 Re: Multiplatform LS : Lord Spigol" 22 Re: GLIBC : John Poltorak 23 RE: wxWindows : Dave Webster 24 RE: wxWindows : Dave Webster 25 Specifying YACC as preference over BISON : John Poltorak 26 Re: Multiplatform LS : Lord Spigol" 27 Re: baseline perl harness results : Maynard" 28 Re: UnixOS/2 bootstrap : Maynard" 29 Re: UnixOS/2 bootstrap : Dave Saville" 30 Re: UnixOS/2 bootstrap : Dave Saville" 31 Re: UnixOS/2 bootstrap : Dave Saville" 32 Re: UnixOS/2 bootstrap : Dave Saville" 33 Re: GCC =?us-ascii?q?3.2=3F?= : Mika Kristian Laitio 34 Re: UnixOS/2 bootstrap : John Poltorak 35 Re: baseline perl harness results : Maynard" 36 Re: GLIBC : Tobias Huerlimann" 37 Re: UnixOS/2 bootstrap : Maynard" 38 Re: UnixOS/2 bootstrap : Dave Saville" 39 Re: wxWindows : John Poltorak 40 Re: GLIBC : Holger Veit 41 Re: Multiplatform LS : John Poltorak 42 Re: Baseline EMX [emx_inst.cmd] : Maynard" 43 Re: baseline perl harness results : Maynard" 44 RE: wxWindows : Dave Webster 45 Re: wxWindows : John Poltorak 46 Re: Multiplatform LS : John Poltorak 47 Re: GLIBC : John Poltorak 48 Re: bootstrap.lst : John Poltorak 49 RE: wxWindows : Dave Webster 50 RE: Multiplatform LS : Lord Spigol" 51 Re: wxWindows : Henry Sobotka 52 Re: UnixOS/2 bootstrap : Dave Saville" 53 RE: wxWindows : Stefan Neis 54 RE: wxWindows : Stefan Neis 55 Re: baseline perl harness results : Stefan Neis 56 Re: baseline perl harness results : John Poltorak 57 Re: UnixOS/2 bootstrap : John Poltorak 58 Re: Baseline EMX [emx_inst.cmd] : John Poltorak 59 Re: baseline perl harness results : John Poltorak 60 Re: wxWindows : John Poltorak 61 Re: UnixOS/2 bootstrap : Dave Saville" 62 RE: wxWindows : Csaba" 63 Re: wxWindows : Csaba" 64 Re: UnixOS/2 bootstrap : Csaba" 65 Re: bootstrap.lst : Csaba" 66 Re: bootstrap.lst : Csaba" **= Email 1 ==========================** Date: Fri, 26 Jul 2002 06:47:02 -0500 (CDT) From: "Maynard" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 11:42:43 +0100, John Poltorak wrote: >> Its not creating the lib dir. Reason is because I have a .wgetrc in >> $HOME that has >> dirstruct = off > > >Is there any WGET option to ignore WGETRC? as I garner from the wget documentation, there are two options: 1. temporarily store, clear, and subsequently restore $HOME so that .wgetrc is not found; 2. clear or restate all critical options on the command line; I don't see how to clear *_everything_*, just the individual options. 3. in this specific case, using -x on the commandline is probably the best solution to test first. [should force the use of directory structure which is certainly the intent; and override .wgetrc] [good eye, Dave!] --Maynard **= Email 2 ==========================** Date: Fri, 26 Jul 2002 07:43:43 -0700 From: Neil Waldhauer Subject: Re: wxWindows On Fri, 26 Jul 2002 09:04:19 -0500, Dave Webster wrote: > Actually, I am looking forward to WatcomC++ on OS/2. I thought it was > already available and I hadn't yet taken the plunge. I understand it comes > with complete support for STL??? I'm now learning ANSI/ISO C++, and compiling examples on OS/2. My textbook is Accelerated C++ by Koenig and Moo. The examples for chapter 00 and 01 work in OpenWatcom if I download and install STLPort. Chapter 02 won't work because Watcom support for namespace isn't quite right. VisualAge C++ v3.08 also works for chapter 00 and 01, but not chapter 02, again after installing STLPort, but that is because VAC 3.08 doesn't have the 'using' keyword. VisualAge 4.0 supports through chapter 04 and I'll have more to say when I finish the book. I'm looking forward to trying wxWindows on OS/2. Neil -- Neil Waldhauer, neil at blondeguy.com My karma ran over my dogma. **= Email 3 ==========================** Date: Fri, 26 Jul 2002 07:51:35 -0500 From: Dave Webster Subject: RE: wxWindows Do you get them with source code and do they support as many compilers per platform as wxWindows does? Note that you just can't share compiled C++ code across different compilers while this works (almost) without problems for all C libraries. Yes, RW (not a very viable toolkit anymore with stl these days) shipped with a very wide array of compiler support. It's Unix dist was traditional Unix, but its Windows-OS/2 dist was very traditional windows with an InstallShield/OS-2 installer type installation. Just clikced the SETUP icon and you were off. Once the files were in place you simply did a make with your appropriate compiler's makefile, but I always found the libs to be fine as they were. And I recently got Cisco's ICM Telephony API toolkit and guess what? Popped the CD in, the Windows Installer came up, I clicked on the license agreement, clicked the "Total" install and "Next" and bang! Complete installation, complete with all source files, headers, libs (import libs and dlls) for all the major Win32 compilers, and samples all with make files. Didn't have to compile, or "make" a thing, it was all right there. That's what Windows and OS/2 users expect. ********************************************************************** So again, I think you got your division of the world in two part wrong, it's not Unix vs. Windows/OS2, it's open source vs. closed source and partially C++ vs. C (and maybe even cross-platform vs. single platform). Well, Open Source, being mostly dominated, especially in the earlier days by Linux based stuff grew up with a Unix-like flavor. And that is yet another reason why Open Source is not as widely penetrating the market as it probably could have. Windows (and OS/2) users expect things to come to them in a traditional Windows-OS2 way, even toolkits. Give me a setup icon to click, let me answer a few environment questions and give me app, complete with start-menu and desktop/WPS icon and, in the case of a toolkit, a makefile for my compiler. That's it. No scripts to hack through no Unix-like silliness to deal with. When Open Source developers come to understand, Open Source or not, they need to look like a traditional commercial package, they'll get the next big increment of market penetration. Until then, they won't. Same with Linux. Linux has to draw it's knew users from Windows, and when the environment behaves more like Windows and less like Unix, it might start to take off a lot more. People will use what they are comfortable with, regardless of whether or not its better. ******************************************************************* Basically it's some clever stuff to autodetect your compiler and some internal details of your platform to automatically build Makefiles from the slightly more abstract Makefile.in that are suitable for your environment. _Very_ convenient, if you want to support more than one compiler on one specific platform. ;-) I seriously doubt configure could generate the VA makefile mainly because it is a very odd one due to the limitations of old VA's make with static builds. Besides, VA users are much more familiar with VA's own MakeMake than autoconf and autoconf is yet another learning curve (that goes for me, too, and I already have a enough learning curves to live with and the older I get the fewer I am willing to tolerate). However, I certainly do not mind if someone else wants to set up the autoconf stuff for wxOS2, but I'm not going to fool with it. **= Email 4 ==========================** Date: Fri, 26 Jul 2002 07:53:48 -0500 (CDT) From: "Maynard" Subject: bootstrap.lst John, I think that wget will maintain a connection over multiple files from the same source, at least under some circumstances. If so, then if bootstrap.lst were sorted ;-} ..... While looking at bootstrap.lst in this regard, I note that my latest version received today, stamped 7-24-02 23:23, has a bad line in it: x04.zip without source Later, --Maynard **= Email 5 ==========================** Date: Fri, 26 Jul 2002 08:14:25 -0400 (EDT) From: "Thomas E. Dickey" Subject: Re: Specifying YACC as preference over BISON On Fri, 26 Jul 2002, John Poltorak wrote: > Is there any way to globally specify a preference for YACC over BISON when > building apps? > > A number of configure scripts search for bison and then byacc, so I guess > the selection is made on the search sequence. > > I wonder if there was something which could be set in CONFIG_SITE along > the lines of USE_YACC_FIRST... set the environment variable 'YACC' to what you're preferring. -- T.E.Dickey http://invisible-island.net ftp://invisible-island.net **= Email 6 ==========================** Date: Fri, 26 Jul 2002 09:04:19 -0500 From: Dave Webster Subject: RE: wxWindows Actually, I am looking forward to WatcomC++ on OS/2. I thought it was already available and I hadn't yet taken the plunge. I understand it comes with complete support for STL??? As for autoconf, just chalk this up to something I would like to have help with in getting wxOS2 finished. I honestly do not have the time to configure that AND get the last bit of coding done at the same time. Getting OS/2 code done always trumps everything else. The main point is that even though toolkits like wxWindows that offer cross platform development, does NOT mean that the resultant applications have to install the same way. That is actually a horrible idea. Applications should install natively, the way users on each platform expect applications to be delivered to them. In fact, wxWindows itself, prides itself on being the only toolkit that delivers NATIVE widgets on all the platforms it supports. Windows apps look like Windows apps, OS/2 apps look like OS/2 apps and so on, not like generic Java junk. By the same token, installations should always be made native. Sometimes open source developers forget they are not open source users. Trust me, the best way to guarantee wide spread rejection of a consumer targeted Open Source Windows-OS/2 application to make a user do "autoconf, configure, make, make install". That's the quickest route to the Recycle Box or Shredder. All I want as a Windows or OS/2 user is the classic Setup.exe or Install.bat icon, nothing else will works if you want wide spread acceptance. Far far too many Open Source zealots just don't get that. They (the users) are NOT us (the developers). -----Original Message----- From: John Poltorak [mailto:jp at eyup.org] Sent: Friday, July 26, 2002 8:34 AM To: os2-unix at eyup.org Subject: Re: wxWindows On Thu, Jul 25, 2002 at 03:33:11PM -0500, Dave Webster wrote: > Same for commercial toolkits. When you install Rogue Wave on OS/2 or > Windows you run the same kind of installer as you do for app. You tell it > where you want it, it puts it there, libs, dlls, headers, source and all. > No configure, no make, not make install, no Unix-like crappola at all. That > is how wxOS2 will install, hopefully. You can build the library if you want > to, but it will have the static lib and dll already built on the CD if I > have any say so in the matter. I have no plans to even mess with this > autoconf/configure stuff, hell I don't even know what it is! Actually, no one knows what it is :-)... it just works. (OK there are one or two people who actually understand what is going on, but in reality there is no need to do so.) Just run:- autoconf configure make make test make install There is a magic suite of tools which just works. We are almost in a position now on OS/2 to enjoy the fruits of this development. You need to broaden your outlook and stop dismissing the Unix 'crappola'. It is designed for cross-platform use. You seem to have a narrow focus on VAC, which is already out of support from IBM, so has no long term future. In contrast OpenWatcom has only just been released and may provide stiff competition to gcc, especially in open source projects over many years to come. By exploring the benefits offered by autoconf+friends you may be able to provide support for gcc and eventually Watcom without needing to do anything special. Once there is a release of which works on OS/2, I'm sure people will see how far things get using autoconf etc. Just think of it as a set of tools for creating makefiles, libraries etc, automatically. The onus is on the wxWindows developer to create the initial configure.in and makefile.in files and everything else should fall into place. I guess we just need to make sure that whatever is done for Linux is sufficiently cross platform to work on OS/2 as well. -- John **= Email 7 ==========================** Date: Fri, 26 Jul 2002 09:32:54 -0500 From: Dave Webster Subject: RE: Multiplatform LS By the same token is there a decent OS/2 cmd processor out there anywhere. I've tried Hobbes and many other places. I really would like to get an OS/2 command line window that I could adjust to something like a 72 x 132 with full font and color selection like the one on Windows 2000 but I have yet to find anything decent. -----Original Message----- From: Lord Spigol [mailto:spigol at hotpop.com] Sent: Friday, July 26, 2002 9:02 AM To: unixos2 Subject: Multiplatform LS Howdy UnixOS2ers There are some place on the web with multiplatform LS? Or then someone can teach me how to compile ls? :) []s Rod **= Email 8 ==========================** Date: Fri, 26 Jul 2002 09:42:23 +0100 From: John Poltorak Subject: GLIBC We have had a number of discussions about GLIBC over time and the general consensus is to avoid it, but where do I get the BSD alternative? I know GLIBC contains regex.c and was hoping I could find the BSD equivalent but I've never been able to locate it. -- John **= Email 9 ==========================** Date: Fri, 26 Jul 2002 10:10:19 +0200 (CEST) From: Stefan Neis Subject: Re: ln.cmd On Thu, 25 Jul 2002, John Poltorak wrote: > at echo off > cp %1 %2 > > I realise that creating the file is > simple enough, but I would like it to have a specific timestamp, so that > it isn't highlighted as a file differing between different locations if a > comparison is ever made. Well, reusing the same time stamp is exactly what the '-p' option does ... Anyway, how do you compare files, if it's fooled into believing in different files just based on date/time? > BTW is 'cp' OK, or should it be 'cp -p'? Well, your description above does sound like you want 'cp -p' ... Regards, Stefan **= Email 10 ==========================** Date: Fri, 26 Jul 2002 10:34:02 -0400 From: Jack Troughton Subject: Re: wxWindows Dave Webster wrote: > Actually, I am looking forward to WatcomC++ on OS/2. I thought it was > already available and I hadn't yet taken the plunge. I understand it comes > with complete support for STL??? It's available... you can get it from ftp.openwatcom.org. While you're at it, get perforce from www.perforce.com and you can get all the sources as well. If you're any good at dealing with make files, I'm helping fix them up so that a developer can get the toolkit and sources and build it out of the box... which means I'm learning wmake these days:) > As for autoconf, just chalk this up to something I would like to have help > with in getting wxOS2 finished. I honestly do not have the time to > configure that AND get the last bit of coding done at the same time. > Getting OS/2 code done always trumps everything else. > > The main point is that even though toolkits like wxWindows that offer cross > platform development, does NOT mean that the resultant applications have to > install the same way. That is actually a horrible idea. Applications > should install natively, the way users on each platform expect applications > to be delivered to them. In fact, wxWindows itself, prides itself on being > the only toolkit that delivers NATIVE widgets on all the platforms it > supports. Windows apps look like Windows apps, OS/2 apps look like OS/2 > apps and so on, not like generic Java junk. By the same token, > installations should always be made native. Sometimes open source > developers forget they are not open source users. > > Trust me, the best way to guarantee wide spread rejection of a consumer > targeted Open Source Windows-OS/2 application to make a user do "autoconf, > configure, make, make install". That's the quickest route to the Recycle > Box or Shredder. All I want as a Windows or OS/2 user is the classic > Setup.exe or Install.bat icon, nothing else will works if you want wide > spread acceptance. Far far too many Open Source zealots just don't get that. > They (the users) are NOT us (the developers). This is true... however, I do think that including the ability to do that with the source release of a given project is a good idea. Basically, if a developer is making an opensource project, they should make a binary release that uses the normal installation methods, and a source release that includes the autoconf yadda stuff. -- ------------------------------------------------------------------- * Jack Troughton jake at consultron.ca * * http://consultron.ca irc.ecomstation.ca * * Laval Qu‚bec Canada news://news.consultron.ca * ------------------------------------------------------------------- **= Email 11 ==========================** Date: Fri, 26 Jul 2002 10:35:44 -0500 (CDT) From: "Maynard" Subject: baseline perl harness results John, thanks again for an ongoing succesful baseline build_system. I just ran it unmodified, first into an existing %uxrt% and then into an empty one. I do find that perl harness can repeatedly give varying answers. Below is my latest and greatest; I can only suggest that when somebody runs the baseline build and receives >12 failed subtests that from the same CLI as used for the basebuild, go to \UnixOS2\workdir\Perl-5.8.0 and run "perl t/harness" with optional logging and see if the answer changes. I cannot speculate as to WHY the results differ; they may even differ if you run harness from .../t/, though I can understand that could be result of relative paths in path or libpath. --Maynard Failed 3/726 test scripts, 99.59% okay. 5/68656 subtests failed, 99.99% okay. Failed Test Stat Wstat Total Fail Failed List of Failed ----------------------------------------------------------------------------- ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 ../lib/Net/t/hostname.t 2 1 50.00% 1 lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 63 tests and 563 subtests skipped. **= Email 12 ==========================** Date: Fri, 26 Jul 2002 10:46:24 +0300 From: Mika Kristian Laitio Subject: GCC =?iso-8859-1?q?3.2=3F?= Do you happen to know has anybody from OS/2 developers worked with the GCC recently? I know that there is GCC 3.0.3 beta but it has been released quite a long time ago and is not widely adopted. Current stable release is GCC 3.1.1 but it seems that at least in the Linux world many companies are planning to skip it and jump directly from gcc 2.9.x to the GCC 3.2 which should be released in a week or two. (Linux Mandrake released just the first 9.0 beta with GCC 3.1 but are now building all packages with GCC 3.2 in order to release next beta with that in the next week) The reason why I am asking this is that the C++ ABI is not binary compatible between GCC 2.9.x <-> GCC 3.1 <-> GCC 3.2... Here is part of the GCC release plan crapped from http://gcc.gnu.org/develop.html +-- GCC 3.0 branch created ------+ | (Feb 12 2001) \ | v v GCC 3.0 release (Jun 18 2001) New development plan announced \ | (Jul 18 2001) v | GCC 3.0.1 release (Aug 20 2001) | \ v v GCC 3.1 Stage 1 (ended Oct 15 2001) GCC 3.0.2 release (Oct 25 2001) | \ v v GCC 3.1 Stage 2 (ended Dec 19 2001) GCC 3.0.3 release (Dec 20 2001) | \ v v GCC 3.1 Stage 3 (ends Feb 15 2002) GCC 3.0.4 release (Feb 20 2001) | +-- GCC 3.1 branch created ------+ | \ | v v GCC 3.1 release (May 15 2002) GCC 3.2 Stage 1 (ends Jun 15 2002) \ | v | GCC 3.1.1 release (Jul 21 2002) | \ v v New development plan announced Branch renamed to GCC 3.2 to | (Jul 14 2002) accomodate for C++ ABI fixes | (C++ binary incompatible with | GCC 3.1, see release info) | \ | v | GCC 3.2 release (Jul 23 2002) | \ | v | GCC 3.2.1 release (Sep 15 2002) | \ | v | GCC 3.2.2 release (Nov 15 2002) v GCC 3.3 Stage 2 (ends Aug 15 2002) Mika **= Email 13 ==========================** Date: Fri, 26 Jul 2002 10:52:24 +0200 (CEST) From: Stefan Neis Subject: Re: GLIBC On Fri, 26 Jul 2002, John Poltorak wrote: > > > We have had a number of discussions about GLIBC over time and the general > consensus is to avoid it, but where do I get the BSD alternative? Holger's EMU - once it's done ... I know, this answer doesn't really help ... Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 14 ==========================** Date: Fri, 26 Jul 2002 10:55:10 +0200 (CEST) From: Stefan Neis Subject: Re: GCC =?iso-8859-1?q?3.2=3F?= On Fri, 26 Jul 2002, Mika Kristian Laitio wrote: > The reason why I am asking this is that the C++ ABI is not binary > compatible between GCC 2.9.x <-> GCC 3.1 <-> GCC 3.2... I knew that gcc-2.7,2.8, 2.95 is incompatible with any 3.x, but that those 3.x versions are being incompatible to each other is news to me - and very bad news at that. You're really sure about this one? Regards, Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 15 ==========================** Date: Fri, 26 Jul 2002 11:01:36 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap On Thu, 25 Jul 2002 18:45:28 +0100, John Poltorak wrote: >On Thu, Jul 25, 2002 at 05:56:29PM +0100, Dave Saville wrote: >> John >> >> Just downloaded and tried your script. > >The script is work-in-progress and is liable to change every day. > >Without knowing what you actually ran I can't say what may have happened. > > >> I used a drive that only had >> /tmp on it unixwise. >> >> first thing was the script ends up going to >> >> cd \%bld_home%\lib >> ux2_inst >> but there is no \lib everything is in the base. > >There shouldn't be a \lib, there should be \%bld_home%\lib. If there isn't >then WGET has not retrieved everything properly. There is no \%bld_home%\lib, ux2_inst is in \%bld_home%\ along with a lot of other stuff. Its not creating the lib dir. Reason is because I have a .wgetrc in $HOME that has dirstruct = off Suggest you either unset $HOME (can you on OS/2?) or add -x to the wget parms. I have just rerun with $HOME set as per normal and -x and lib is created & populated. Gonna let it run now and see how far I get this time. -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 16 ==========================** Date: Fri, 26 Jul 2002 11:02:10 -0300 (ADT) From: "Lord Spigol" Subject: Multiplatform LS Howdy UnixOS2ers There are some place on the web with multiplatform LS? Or then someone can teach me how to compile ls? :) []s Rod **= Email 17 ==========================** Date: Fri, 26 Jul 2002 11:08:31 +0200 (CEST) From: Stefan Neis Subject: RE: wxWindows On Thu, 25 Jul 2002, Dave Webster wrote: > And not to be condescending, but this where Unix/Linux folks just don't get > it when dealing with OS/2 and Windows. There is no autoconf, configure, > make, make install for Windows or OS/2 software, at least not in the > traditional sense. To me, that always rather seemed a question of whether you're distributing source code or binaries. Take e.g. Netscape/StarOffice/commercial Unices on the one hand (no configure;make; make install - not even on Linux) and e.g. emacs on the other hand (source code even on OS/2, Windows (well, there are precompiled binaries for many of the more popular OSes like Solaris, Windows, OS/2, Linux, but the "original" package is source code only)). > Same for commercial toolkits. When you install Rogue Wave on OS/2 or > Windows you run the same kind of installer as you do for app. Do you get them with source code and do they support as many compilers per platform as wxWindows does? Note that you just can't share compiled C++ code across different compilers while this works (almost) without problems for all C libraries. So again, I think you got your division of the world in two part wrong, it's not Unix vs. Windows/OS2, it's open source vs. closed source and partially C++ vs. C (and maybe even cross-platform vs. single platform). > I have no plans to even mess with this > autoconf/configure stuff, hell I don't even know what it is! Basically it's some clever stuff to autodetect your compiler and some internal details of your platform to automatically build Makefiles from the slightly more abstract Makefile.in that are suitable for your environment. _Very_ convenient, if you want to support more than one compiler on one specific platform. ;-) Regards, Stefan **= Email 18 ==========================** Date: Fri, 26 Jul 2002 11:14:28 -0500 (CDT) From: "Maynard" Subject: Baseline EMX [emx_inst.cmd] John, Part of the emx construction seeks to move /emx/lib/* to /usr/lib/ by using 'mv' When the target directory /usr/lib already exists (even empty) the move fails mv: usr/lib: cannot overwrite directory and consequently the baseline is incorrect for future anticipations, adding about a dozen failures to the harness test, perhaps most notably; ../lib/ExtUtils/t/Embed.t 9 9 100.00% 1-9 Since this move entails subdirectories of /emx/lib, a simple 'copy' cannot be substituted; erasing the target first would be rude; 'xcopy' could be used, followed by 'delete'. While solving this problem, lets [re]consider the use of moving /emx to /usr I will offer that emx is of enough significance (as is perl) to merit the path of its own; but most particularly if it is likely to be updated, as it would seem then to be most fortunate to ensure no straggling relics. Downside of leaving /emx ? one additional target for path and libpath ? -- Maynard **= Email 19 ==========================** Date: Fri, 26 Jul 2002 11:36:07 -0500 (CDT) From: "Maynard" Subject: Re: bootstrap.lst On Fri, 26 Jul 2002 16:38:31 +0100, John Poltorak wrote: >> I think that wget will maintain a connection over multiple files from >> the same source, at least under some circumstances. If so, then if >> bootstrap.lst were sorted ;-} ..... > >Would sorting it make it more efficient? wget has a "keep-alive" feature for persistent connections; however it may only apply to http connections; I haven't figured it out completely yet. Much of the baseline list is http anyway so it should work. I recall seeing it work while actually getting the files; now that the sources are not updated, the process appears to be making new connections for each file -- Maynard **= Email 20 ==========================** Date: Fri, 26 Jul 2002 11:42:43 +0100 From: John Poltorak Subject: Re: UnixOS/2 bootstrap On Fri, Jul 26, 2002 at 11:01:36AM +0100, Dave Saville wrote: > On Thu, 25 Jul 2002 18:45:28 +0100, John Poltorak wrote: > > >> first thing was the script ends up going to > >> > >> cd \%bld_home%\lib > >> ux2_inst > >> but there is no \lib everything is in the base. > > > >There shouldn't be a \lib, there should be \%bld_home%\lib. If there isn't > >then WGET has not retrieved everything properly. > > There is no \%bld_home%\lib, ux2_inst is in \%bld_home%\ along with > a lot of other stuff. > > Its not creating the lib dir. Reason is because I have a .wgetrc in > $HOME that has > dirstruct = off Is there any WGET option to ignore WGETRC? > Suggest you either unset $HOME (can you on OS/2?) or add -x to the > wget parms. > > I have just rerun with $HOME set as per normal and -x and lib is > created & populated. Gonna let it run now and see how far I get this > time. The baseline environment only takes a couple of minutes to set up, and you can normally tell if Perl has any chance of succeeding as it falls over fairly quickly if the environment has something missing. > -- > Regards > > Dave Saville > Please note new email address dave.saville at ntlworld.com > -- John **= Email 21 ==========================** Date: Fri, 26 Jul 2002 11:54:02 -0300 (ADT) From: "Lord Spigol" Subject: Re: Multiplatform LS Exactly what I need to compile? There is an Unix environment package for OS2 on hobbes called gnupack.wpi. Elsewhere there is the GCC on WPI also. I feel a bit losty. :) []s Rod On Fri, 26 Jul 2002 15:14:46 +0100, John Poltorak wrote: >make -f makefile.os2 **= Email 22 ==========================** Date: Fri, 26 Jul 2002 12:10:21 +0100 From: John Poltorak Subject: Re: GLIBC On Fri, Jul 26, 2002 at 10:52:24AM +0200, Stefan Neis wrote: > On Fri, 26 Jul 2002, John Poltorak wrote: > > > > > > > We have had a number of discussions about GLIBC over time and the general > > consensus is to avoid it, but where do I get the BSD alternative? > > Holger's EMU - once it's done ... OK, but what is Holger's EMU based on? Does the Posix/2 archive detail where the original source comes from? All I'm looking for is a source for regex.c. There is one in GNU libc, but it doesn't seem right to have to download a 16MB archive just to grab a 250kB file... > Stefan > -- > Micro$oft is not an answer. It is a question. The answer is 'no'. > -- John **= Email 23 ==========================** Date: Fri, 26 Jul 2002 12:15:01 -0500 From: Dave Webster Subject: RE: wxWindows Just one more ??? on autoconf. If autoconf generates makefiles with POSIX flags or requires "makes" to work with POSIX flags in makefiles, and POSIX is a Unix specification what makes it so great as a "cross-platform" tool for builds? Seems more like it forces everyone to be UNIX-like, gag.... Seems to me for an auto makefile generator, it should be able to generate a makefile that is "normal" for the target, in otherwords, if Watcom for Windows or OS/2 does not support POSIX flags in makefile (and there is no compelling reason why they should) for a make generator to truely be a good cross-platform tool it should be able to generate a makefile with or without some OS-specific flags in it..... **= Email 24 ==========================** Date: Fri, 26 Jul 2002 12:34:34 -0500 From: Dave Webster Subject: RE: wxWindows Agree for the most part. Seems like a LOT of work to get a non-supported compiler into this environment. If all you are looking for is a cross-platform "nmake", Opus is as good as I've found anywhere (and I'm not one for whom EVERYTHING has to be free). And as for autogenerating make environments (particular build dependencies) my experience is that most commercial software firms and larger companies with internal development staffs have their own environment generators anyway. Think for a commercial firm, anyway, using a commercial product like Opus make, which supports just about everything in existence is much less painful, even if you have to fork a few $$$'s (US or otherwise) to get it. Personally all this effort that goes into wx's configure stuff, a tmake to generate a config script and then running an autoconf against that just to get a makefile....classic lack of focus on the real issue, the library itself. I know they started this because they got tired of maintaining so many makefiles, but to me, if there isn't enough interest in the user community for a particular compiler for some who uses that compiler every day to supply the makefile and keep it up to date, then the compiler is probably not worth supporting in the first place! It's really a lack of focus issue so prevalent in open source efforts. -----Original Message----- From: Stefan Neis [mailto:neis at cdc.informatik.tu-darmstadt.de] Sent: Friday, July 26, 2002 11:58 AM To: 'os2-unix at eyup.org' Subject: RE: wxWindows On Fri, 26 Jul 2002, Dave Webster wrote: > When Open Source developers come to > understand, Open Source or not, they need to look like a traditional > commercial package, they'll get the next big increment of market > penetration. Until then, they won't. I think nobody is going to doubt those facts for the typical end-user application. For packages targetting developpers (especially cross-platform), things are slightly different. Personally, I tend to get irritated if I have to follow one particular installation scheme on Windows, another one on OS/2 and a third one on Linux for the same application/library. > I seriously doubt configure could generate the VA makefile mainly because it > is a very odd one due to the limitations of old VA's make with static > builds. So do I. The first thing is that all the required knowledge about a compilation environment has to have been worked into autoconf _first_. I seem to recall that some VA support was incorporated for building Mozilla/Netscape, but that might have been rather limited support. > Besides, VA users are much more familiar with VA's own MakeMake > than autoconf and autoconf is yet another learning curve (that goes for me, > too, and I already have a enough learning curves to live with and the older > I get the fewer I am willing to tolerate). I got familiar enough with autoconf to be able to modify the existing configure.in slightly so it works better with EMX (which basically is gcc, so autoconf/configure does basically know everything that's needed) but writing one from scratch would be way out of my league. Besides, since you have working makefiles, I don't see much point in trying to auto-generate them... The tmake stuff (yet another level of indirection) used by wxWindows to autogenerate the input files for autoconf/configure _and_ the Makefiles for the compilers without autoconf support (VC, Metrowerks, Borland and others) might be more interesting for that purpose - My impression of that system is that it might really be usable for VAC as well (it would essentially imply to strip the file lists out of the existing makefiles and replace them with an 'include' to an autogenerated file containing those file lists. And this autogenerated file essentially already exists for usage by EMX). Regards, Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 25 ==========================** Date: Fri, 26 Jul 2002 12:40:29 +0100 From: John Poltorak Subject: Specifying YACC as preference over BISON Is there any way to globally specify a preference for YACC over BISON when building apps? A number of configure scripts search for bison and then byacc, so I guess the selection is made on the search sequence. I wonder if there was something which could be set in CONFIG_SITE along the lines of USE_YACC_FIRST... Not being very familiar with what either of them do, or what difference there is between them, I noticed when I built FLEX, that parse.c produced by bison was 50k, but only 34k by yacc. I don't know whether to interpret that as bloat on behalf of bison, or missing features in yacc, although the resultant flex.exe was the same size. -- John **= Email 26 ==========================** Date: Fri, 26 Jul 2002 12:54:36 -0300 (ADT) From: "Lord Spigol" Subject: Re: Multiplatform LS Yes, but I dont know what install yet. :( Rod On Fri, 26 Jul 2002 15:59:42 +0100, John Poltorak wrote: >I've spent the last couple of weeks asking people to test a baseline >development environment for building Unix apps on OS/2, which I have been >trying to establish. > >Didn't you read any of the msgs? **= Email 27 ==========================** Date: Fri, 26 Jul 2002 13:01:58 -0500 (CDT) From: "Maynard" Subject: Re: baseline perl harness results Stefan, >I wonder why nobody is talking about those skipped tests. I've been keeping an eye on the number of skipped tests, and would detect an alarm if it were an unusual number; the differences are slight. I also looked into some individual tests and logfiles and concluded that the skips were largely a consequence of OS, hardware, environment and probably pretty accurate; at least I don't feel smart enough to challenge the situation ;-} -- Maynard **= Email 28 ==========================** Date: Fri, 26 Jul 2002 13:04:51 -0500 (CDT) From: "Maynard" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 18:35:41 +0100 (BST), Dave Saville wrote: >I notice that after the build, wget is not in /usr/bin - should it >not be moved/copied from the build side of the fence? I think that it should be copied, but also that it needs to say in /unixos2/lib where it is required in cases where it is not on path; the bootstrap depends upon it. -- Maynard **= Email 29 ==========================** Date: Fri, 26 Jul 2002 13:14:45 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 11:42:43 +0100, John Poltorak wrote: >Is there any WGET option to ignore WGETRC? > > Don't believe so. As you cannot know what someone may or may not have in $HOME I think the safest route would be to create a temp dir and set $HOME to that. That way it can never find any . files for any of the stuff that you might be using during the build. >The baseline environment only takes a couple of minutes to set up, and you >can normally tell if Perl has any chance of succeeding as it falls over >fairly quickly if the environment has something missing. It's happily charging through the tests at present. Just had a few cmd windows open and close on me. There seem to be a lot of "skipping test on this platform" messages. I see that some of them are for the thread stuff - I had to install posix threads for something - can perl not use it? Here are (I think) all of the errors (later no they are not - got fed up copy/pasting em :-)) ----------------------------------------------- const-c.inc:404: warning: assignment makes integer from pointer without a cast (several of these) ---------------------------------------------------- Making POSIX (dynamic) Note (probably harmless): No library found for -lposix Note (probably harmless): No library found for -lcposix ---------------------------------------------------------- t/io/pipe............................Error reading "/no_such_process": No such f ile or directory at io/pipe.t line 186. D:/BIN/sh.exe: /no_such_process: not found ok ------------------------------------------------------------ t/lib/os2_process....................# Failed test (lib/os2_process.t at lin e 252) # got: '2147483885' # expected: '2147484801' # Failed test (lib/os2_process.t at line 301) [Win]QueryWindowTextLength: SYS4097=0x1001: [No description found in OSO001.MSG] at lib/os2_process.t line 308. # Looks like you planned 227 tests but only ran 106. # Looks like your test died just after 106. FAILED at test 90 t/lib/os2_process_kid................# Failed test (lib/os2_process.t at lin e 252)^M # got: '2147483885'^M # expected: '2147483772'^M Process terminated by SIGPIPE ^M Process terminated by SIGTERM^M FAILED at test 90 -------------------------------------------------------------------- lib/ExtUtils/t/basic.................make[5]: *** No rule to make target `../../ ../config.sh', needed by `../../../lib/Config.pm'. Stop. make[4]: *** [subdirs] Error 2 make[3]: *** [disttest] Error 2 # Failed test (../lib/ExtUtils/t/basic.t at line 108) # got: '512' # expected: '0' ---------------------------------------------------------------------- # make[3]: Leaving directory `/unixos2/workdir/perl-5.8.0/t/Big-Dummy' Process terminated by SIGPIPE FAILED at test 14 ---------------------------------------------------------------------- --- Failed 4 test scripts out of 660, 99.39% okay. ### Since not all tests were successful, you may want to run some of ### them individually and examine any diagnostic messages they produce. ### See the INSTALL document's section on "make test". ### You have a good chance to get more information by running ### ./perl harness ### in the 't' directory since most (>=80%) of the tests succeeded. u=1297.96 s=0 cu=0 cs=0 scripts=660 tests=68646 make[2]: *** [_test_tty] Error 1 make[2]: Leaving directory `/unixos2/workdir/perl-5.8.0' make[1]: *** [_test] Error 2 make[1]: Leaving directory `/unixos2/workdir/perl-5.8.0' make: *** [test] Error 2 ---------------------------------------------------------------------- ----------- e:/usr/lib/perl/lib/5.8.0/Attribute/Handlers/demo/Demo.pm e:/usr/lib/perl/liThe link function is unimplemented at installperl line 684. b/5.8.0/Attribute/Handlers/demo/demo2.pl e:/usr/lib/perl/lib/5.8.0/Attribute/Handlers/demo/demo3.pl ---------------------------------------------------------------------- -------------- e:/usr/lThe link function is unimplemented at installperl line 684. The link function is unimplemented at installperl line 684. ib/perl/bin/perlcc Get several of these ---------------------------------------------------------------------- --------------- Making DynaLoader (static_pic) make[3]: Entering directory `/unixos2/workdir/perl-5.8.0/ext/DynaLoader' make[3]: Leaving directory `/unixos2/workdir/perl-5.8.0/ext/DynaLoader' make[3]: Entering directory `/unixos2/workdir/perl-5.8.0/ext/DynaLoader' make[3]: Leaving directory `/unixos2/workdir/perl-5.8.0/ext/DynaLoader' e:/usr/bin/sh.exe: e:\emx\bin: cannot execute - Is a directory e:/usr/bin/sh.exe: e:\usr\local\bin: not found e:/usr/bin/sh.exe: c:\os2: cannot execute - Is a directory cat: extras.lst: No such file or directory make[2]: Leaving directory `/unixos2/workdir/perl-5.8.0' LD_LIBRARY_PATH=/unixos2/workdir/perl-5.8.0 ./perl installman ---------------------------------------------------------------------- --------------------- op/stat................................# Failed at op/stat.t line 318 FAILED test 44 Failed 1/73 tests, 98.63% okay (less 28 skipped tests: 44 okay, 60.27%) ---------------------------------------------------------------------- -------------------------------------------- lib/os2_process........................# Failed test (lib/os2_process.t at l ine 252) # got: '2147483885' # expected: '2147484510' # Failed test (lib/os2_process.t at line 301) [Win]QueryWindowTextLength: SYS4097=0x1001: [No description found in OSO001.MSG] at lib/os2_process.t line 308. # Looks like you planned 227 tests but only ran 106. # Looks like your test died just after 106. dubious Test returned status 255 (wstat 65280, 0xff00) DIED. FAILED tests 90, 106-227 Failed 123/227 tests, 45.81% okay lib/os2_process_kid....................# Failed test (lib/os2_process.t at l ine 252)^M # got: '2147483885'^M # expected: '2147484385'^M # Failed test (lib/os2_process.t at line 407)^M # Structures begin differing at:^M # $got->[0] = '276'^M # $expected->[0] = '-4'^M # Failed test (lib/os2_process.t at line 452)^M # Structures begin differing at:^M # $got->[0] = '276'^M # $expected->[0] = '-4'^M # Looks like you failed 3 tests of 227.^M FAILED tests 90, 174, 209 Failed 3/227 tests, 98.68% okay ------------------------------------------------------------- lib/rx_dllld...........................skipped all skipped: cannot find RXU.DLL lib/rx_emxrv...........................ok lib/rx_objcall.........................skipped all skipped: cannot find RXU.DLL lib/rx_tievar..........................skipped all skipped: cannot find RXU.DLL lib/rx_tieydb..........................skipped all skipped: cannot find RXU.DLL lib/rx_varset..........................ok lib/rx_vrexx...........................skipped all skipped: cannot find VREXX.DLL ------------------------------------------------------------------ ../lib/ExtUtils/t/basic................make[2]: *** No rule to make target `../. ./../config.sh', needed by `../../../lib/Config.pm'. Stop. make[1]: *** [subdirs] Error 2 make: *** [disttest] Error 2 # Failed test (../lib/ExtUtils/t/basic.t at line 108) # got: '512' # expected: '0' -------------------------------------------------------------------- Its still throwing errors - like network which claims no loopback which is crap. How long does this thing take? - its been running since 10:00 on a 400Mhz box with oodles of memory :-) -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 30 ==========================** Date: Fri, 26 Jul 2002 13:17:44 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap 'nother thing. The perl tests seem to be run twice. The first time thru a test will fail and a reason given. The second the same test will fail and it says "no reason given" ? -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 31 ==========================** Date: Fri, 26 Jul 2002 13:21:48 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap Its finally finished :-) Failed 4/726 test scripts, 99.45% okay. 128/68656 subtests failed, 99.81% okay. Failed Test Stat Wstat Total Fail Failed List of Failed ---------------------------------------------------------------------- --------- ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 lib/os2_process.t 255 65280 227 123 54.19% 90 106-227 lib/os2_process_kid.t 227 3 1.32% 90 174 209 op/stat.t 73 1 1.37% 44 63 tests and 551 subtests skipped. -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 32 ==========================** Date: Fri, 26 Jul 2002 13:24:07 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 06:47:02 -0500 (CDT), Maynard wrote: >On Fri, 26 Jul 2002 11:42:43 +0100, John Poltorak wrote: > >>> Its not creating the lib dir. Reason is because I have a .wgetrc in >>> $HOME that has >>> dirstruct = off >> >> >>Is there any WGET option to ignore WGETRC? > >as I garner from the wget documentation, there are two options: > >1. temporarily store, clear, and subsequently restore $HOME so that >.wgetrc is not found; > >2. clear or restate all critical options on the command line; I don't >see how to clear *_everything_*, just the individual options. > >3. in this specific case, using -x on the commandline is probably the >best solution to test first. [should force the use of directory >structure which is certainly the intent; and override .wgetrc] > >[good eye, Dave!] Thanks - As I said in a previous post, I think the best option is to use a fake $HOME for the duration of the install/builds - There are all sorts of *nixy things that have dot files in $HOME that could throw the build off. -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 33 ==========================** Date: Fri, 26 Jul 2002 13:29:53 +0300 From: Mika Kristian Laitio Subject: Re: GCC =?us-ascii?q?3.2=3F?= Yes, this is from the http://gcc.gnu.org/gcc-3.2/changes.html This GCC release is based on the GCC 3.1 sourcebase, and thus has all the changes in the GCC 3.1 series. In addition, GCC 3.2 has a number of C++ ABI fixes which make its C++ compiler generate binary code which is incompatible with the C++ compilers found in earlier GCC releases, including GCC 3.1 and GCC 3.1.1 In addition following response was sent today to Mandrake beta list by from of the MandrakeSoft's developers: ----------------------------------------------------------- On Thu, 25 Jul 2002, David Walser wrote: > Does this mean we'll be incompatible with RedHat 8.0, > as well as all commercial software that finally does > get recompiled for GCC 3.1? All Linux distributors are switching to gcc3.2, now. Netscape et Mozilla will switch their builds to gcc3.2 as well. It would be surprising we wouldn't use something we requested/suggested for -------------------------------------------------- > > The reason why I am asking this is that the C++ ABI is not binary > > compatible between GCC 2.9.x <-> GCC 3.1 <-> GCC 3.2... > > I knew that gcc-2.7,2.8, 2.95 is incompatible with any 3.x, but > that those 3.x versions are being incompatible to each other is > news to me - and very bad news at that. You're really sure about > this one? **= Email 34 ==========================** Date: Fri, 26 Jul 2002 13:58:04 +0100 From: John Poltorak Subject: Re: UnixOS/2 bootstrap On Fri, Jul 26, 2002 at 01:21:48PM +0100, Dave Saville wrote: > Its finally finished :-) > > Failed 4/726 test scripts, 99.45% okay. 128/68656 subtests failed, > 99.81% okay. > Failed Test Stat Wstat Total Fail Failed List of > Failed > ---------------------------------------------------------------------- > --------- > ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 > lib/os2_process.t 255 65280 227 123 54.19% 90 106-227 > lib/os2_process_kid.t 227 3 1.32% 90 174 209 > op/stat.t 73 1 1.37% 44 > 63 tests and 551 subtests skipped. That looks pretty good. Do you use Object REXX by any chance? You don't get the lib/rx_cmprt.t failure and it has been suggested that this only occurs when people use Classic REXX... > -- > Regards > > Dave Saville > Please note new email address dave.saville at ntlworld.com -- John **= Email 35 ==========================** Date: Fri, 26 Jul 2002 14:01:50 -0500 (CDT) From: "Maynard" Subject: Re: baseline perl harness results On Fri, 26 Jul 2002 19:18:07 +0100, John Poltorak wrote: >If anyone knows how to get a summary of skipped tests, that would be >useful. grep for "skipping" in a log file for 'make test' will show which tests were skipped in a form of lib/Tie/File/t/21_win32..............skipping test on this platform looking at the harness log offers the same information grep -ic \.skipped harness ../lib/Tie/File/t/21_win32.............skipped harness yields other information as well grep skipped: harness | sort -u 9/13 skipped: various reasons all skipped: GDBM_File was not built all skipped: I18N::Langinfo or POSIX unavailable all skipped: IPC::SysV was not built all skipped: Just testing plan & skip_all all skipped: NDBM_File was not built all skipped: Non-Unix platform all skipped: ODBM_File was not built all skipped: This is not BeOS all skipped: This is not NW5 all skipped: This is not VMS all skipped: This is not Win32 all skipped: This is not cygwin all skipped: Your IVs are no larger than your longs all skipped: cannot find RXU.DLL all skipped: cannot find VREXX.DLL all skipped: crypt unimplemented all skipped: network dependent test all skipped: no /etc/group file all skipped: no /etc/passwd file all skipped: no 64-bit file offsets all skipped: no 64-bit types all skipped: no `id` or `groups` all skipped: no ithreads all skipped: no loopback net all skipped: no reason given all skipped: no termcap available to test all skipped: no threads all skipped: no use5005threads all skipped: no useithreads all skipped: not completed **= Email 36 ==========================** Date: Fri, 26 Jul 2002 14:07:31 +0100 (MEZ) From: "Tobias Huerlimann" Subject: Re: GLIBC On Fri, 26 Jul 2002 12:10:21 +0100, John Poltorak wrote: > On Fri, Jul 26, 2002 at 10:52:24AM +0200, Stefan Neis wrote: > > On Fri, 26 Jul 2002, John Poltorak wrote: > > > > > We have had a number of discussions about GLIBC over time and the general > > > consensus is to avoid it, but where do I get the BSD alternative? > > > > Holger's EMU - once it's done ... > > OK, but what is Holger's EMU based on? AFAIK Holger's EMU is based on the libc from FreeBSD: ftp://ftp.freebsd.org/pub/FreeBSD/FreeBSD-stable/src/lib/libc/ Or try the one from OpenBSD: ftp://ftp.openbsd.org/pub/OpenBSD/src/lib/libc/ > All I'm looking for is a source for regex.c. There is one in GNU libc, but > it doesn't seem right to have to download a 16MB archive just to grab a > 250kB file... BSD libc does not have a single regex.c file. Instead, the regular expression engine is split in a few .c files (regcomp.c, engine.c, regexec.c, ...): ftp://ftp.freebsd.org/pub/FreeBSD/FreeBSD-stable/src/lib/libc/regex/ ftp://ftp.openbsd.org/pub/OpenBSD/src/lib/libc/regex/ Porting shouldn't be too difficult but you will have to write your own Makefiles as the Makefile in this directory assumes a complete BSD source tree. I'm not sure if this regex library is the one you're looking for because the BSD libc regex functions might differ from the GNU libc ones. Tobias **= Email 37 ==========================** Date: Fri, 26 Jul 2002 14:08:45 -0500 (CDT) From: "Maynard" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 19:32:19 +0100, John Poltorak wrote: >Ideally, I should use programs which are automatically available to the >user in the first place so it would be better to use FTP for the initial >file retrieval, since FTP is part of a standard OS/2 installation. > >Can anyone provide a simple FTP script? The user also needs: ux2_bootstrap.cmd some general clues or instructions to get started, what to expect for space requirements, etc. Perhaps it is safe to presume that new user has a browser and should be sent to a webpage which offers these bare essentials, including wget -- Maynard **= Email 38 ==========================** Date: Fri, 26 Jul 2002 14:33:41 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 13:58:04 +0100, John Poltorak wrote: >On Fri, Jul 26, 2002 at 01:21:48PM +0100, Dave Saville wrote: >> Its finally finished :-) >> >> Failed 4/726 test scripts, 99.45% okay. 128/68656 subtests failed, >> 99.81% okay. >> Failed Test Stat Wstat Total Fail Failed List of >> Failed >> ---------------------------------------------------------------------- >> --------- >> ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 >> lib/os2_process.t 255 65280 227 123 54.19% 90 106-227 >> lib/os2_process_kid.t 227 3 1.32% 90 174 209 >> op/stat.t 73 1 1.37% 44 >> 63 tests and 551 subtests skipped. > >That looks pretty good. > >Do you use Object REXX by any chance? > Yes >You don't get the lib/rx_cmprt.t failure and it has been suggested that >this only occurs when people use Classic REXX... -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 39 ==========================** Date: Fri, 26 Jul 2002 14:34:17 +0100 From: John Poltorak Subject: Re: wxWindows On Thu, Jul 25, 2002 at 03:33:11PM -0500, Dave Webster wrote: > Same for commercial toolkits. When you install Rogue Wave on OS/2 or > Windows you run the same kind of installer as you do for app. You tell it > where you want it, it puts it there, libs, dlls, headers, source and all. > No configure, no make, not make install, no Unix-like crappola at all. That > is how wxOS2 will install, hopefully. You can build the library if you want > to, but it will have the static lib and dll already built on the CD if I > have any say so in the matter. I have no plans to even mess with this > autoconf/configure stuff, hell I don't even know what it is! Actually, no one knows what it is :-)... it just works. (OK there are one or two people who actually understand what is going on, but in reality there is no need to do so.) Just run:- autoconf configure make make test make install There is a magic suite of tools which just works. We are almost in a position now on OS/2 to enjoy the fruits of this development. You need to broaden your outlook and stop dismissing the Unix 'crappola'. It is designed for cross-platform use. You seem to have a narrow focus on VAC, which is already out of support from IBM, so has no long term future. In contrast OpenWatcom has only just been released and may provide stiff competition to gcc, especially in open source projects over many years to come. By exploring the benefits offered by autoconf+friends you may be able to provide support for gcc and eventually Watcom without needing to do anything special. Once there is a release of which works on OS/2, I'm sure people will see how far things get using autoconf etc. Just think of it as a set of tools for creating makefiles, libraries etc, automatically. The onus is on the wxWindows developer to create the initial configure.in and makefile.in files and everything else should fall into place. I guess we just need to make sure that whatever is done for Linux is sufficiently cross platform to work on OS/2 as well. -- John **= Email 40 ==========================** Date: Fri, 26 Jul 2002 15:10:03 +0200 From: Holger Veit Subject: Re: GLIBC On Fri, Jul 26, 2002 at 02:07:31PM +0100, Tobias Huerlimann wrote: > On Fri, 26 Jul 2002 12:10:21 +0100, John Poltorak wrote: > > > On Fri, Jul 26, 2002 at 10:52:24AM +0200, Stefan Neis wrote: > > > On Fri, 26 Jul 2002, John Poltorak wrote: > > > > > > > We have had a number of discussions about GLIBC over time and the general > > > > consensus is to avoid it, but where do I get the BSD alternative? > > > > > > Holger's EMU - once it's done ... > > > > OK, but what is Holger's EMU based on? > > AFAIK Holger's EMU is based on the libc from FreeBSD: You know more than I... The core part which needs most of my attention is the system calls (the man(2) functions). Once they are available, you can almost use any *BSD libc, and in principle glibc as well (but the interface to integrate a new set of syscalls to glibc is an example of bad design - someone started with it somehow, and others patched the logic to death in an attempt to turn a narrow-minded restricted design into a generally usable one) There are some slight differences between FreeBSD and NetBSD and OpenBSD system call sets, but I ignore these details for now. > Or try the one from OpenBSD: > ftp://ftp.openbsd.org/pub/OpenBSD/src/lib/libc/ It is not a problem to take any of these libs and compile them; part of this has been already demonstrated with the POSIX/2 experiment. The syscall layer is the difficulty. > > All I'm looking for is a source for regex.c. There is one in GNU libc, but > > it doesn't seem right to have to download a 16MB archive just to grab a > > 250kB file... [...] > > I'm not sure if this regex library is the one you're looking for > because the BSD libc regex functions might differ from the GNU libc > ones. There are multiple versions of standalone regex libraries on the net; they usually fall in one of the following two categories. Autoconf'd programs usually can deal with any of them. One style is: re_comp() re_exec() the other is regcomp() regexec() regerror() regfree() You better search for the second one; this is the modern POSIX style. Holger -- Please update your tables to my new e-mail address: holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) **= Email 41 ==========================** Date: Fri, 26 Jul 2002 15:14:46 +0100 From: John Poltorak Subject: Re: Multiplatform LS On Fri, Jul 26, 2002 at 11:02:10AM -0300, Lord Spigol wrote: > Howdy UnixOS2ers > > There are some place on the web with multiplatform LS? > > Or then someone can teach me how to compile ls? :) Grab:- http://hobbes.nmsu.edu/pub/os2/util/file/gnufutil.zip Unzip and then run:- make -f makefile.os2 > []s > Rod -- John **= Email 42 ==========================** Date: Fri, 26 Jul 2002 15:29:07 -0500 (CDT) From: "Maynard" Subject: Re: Baseline EMX [emx_inst.cmd] John, >As it stands the install works fine for a pristine environment, As it stands, and looking forward, this should probably always be the primary objective of this project, creating a useful build environment from scratch, on almost any os/2 installation with the available %bldrt% and %uxrt% partition space. >the complications only arise for a subsequent install. and some unforseen environments, most lately the presence of $HOME/.wgetrc >What I'd like to do is provide an uninstall, and/or option for leaving the >already installed toolset in place. And here's where I'm looking at the project, .... now that I have a good baseline build environment, ...., "what do I do with it?". As answers to that question, - I get rid of the old \emx stuff I've had around from the runtime; - I get rid of the duplicate, and probably conflicting, versions of things like grep and sed which are elsewhere on the path; - I modify config.sys to exclude the old and to include the new; - I probably make up CLIs and WPS objects for the new environment; - I acknowledge that the fine baseline is a minimalist environment, and that I want to add additional "stuff" to %uxrt% - Sooner or later, currently weekly ;-}, you or I are going to expand the definition of "baseline" in this context, and want to add something, and its dependencies. It would be a far far better thing if ux2_inst.cmd could gracefully do this addition, or update another existing element of the baseline package, without having to reformat the partition first ;-} especially if I don't want to recreate my personal extensions. So, if the only end in sight is to create a baseline environment which will build perl, we're done. If the project now becomes one of continuing development of a working build environment, or office environment with an enhanced toolset [%uxrt%\usr\bin in PATH], and with integrated build capacity (no reboot required) then we are well poised to proceed, eh. Thanks again, -- Maynard **= Email 43 ==========================** Date: Fri, 26 Jul 2002 15:39:09 -0500 (CDT) From: "Maynard" Subject: Re: baseline perl harness results On Fri, 26 Jul 2002 20:30:13 +0100, John Poltorak wrote: >I guess we need to decide whether to use the results from 'perl test' or >'perl harness'.... > > >My preference would be 'perl test'. It seems to me an issue of granularity. 'make test' and 'perl harness' should run pretty much the same tests with harness providing more useful detail in the output. or so it seems to me ;-} -- Maynard **= Email 44 ==========================** Date: Fri, 26 Jul 2002 15:40:31 -0500 From: Dave Webster Subject: RE: wxWindows OK, sounds good to me, I had only heard of POSIX used in conjunction with Unix efforts. But so far wxWindows for OS/2 works only under VisualAge and EMX and both have valid and seldom changing makefiles, so I see no need to futz with autoconf at all, especially since it has no hope of working for VA ever or Watcom any time soon. -----Original Message----- From: John Poltorak [mailto:jp at eyup.org] Sent: Friday, July 26, 2002 2:55 PM To: os2-unix at eyup.org Subject: Re: wxWindows On Fri, Jul 26, 2002 at 12:15:01PM -0500, Dave Webster wrote: > Just one more ??? on autoconf. If autoconf generates makefiles with POSIX > flags or requires "makes" to work with POSIX flags in makefiles, and POSIX > is a Unix specification what makes it so great as a "cross-platform" tool > for builds? POSIX is the Portable Operating System Interface specified by the IEEE. It has nothing to do with Unix AFAICT, although many flavours of Unix attempt to comply with POSIX. IMV it is a good thing for Open Source Software to comply with POSIX standards. See:- http://standards.ieee.org/regauth/posix/index.html -- John **= Email 45 ==========================** Date: Fri, 26 Jul 2002 15:49:29 +0100 From: John Poltorak Subject: Re: wxWindows On Fri, Jul 26, 2002 at 09:04:19AM -0500, Dave Webster wrote: > Actually, I am looking forward to WatcomC++ on OS/2. I thought it was > already available and I hadn't yet taken the plunge. It _is_ already available. What I meant was it couldn't be used with autoconf yet because it's make utility does not support POSIX flags which is something that autoconf requires. What it needs is some sort of GNU Make convertor and then it will be usable, supposedly. Apparently it's fairly easy to write one, but no one has bothered to do so yet. > As for autoconf, just chalk this up to something I would like to have help > with in getting wxOS2 finished. I honestly do not have the time to > configure that AND get the last bit of coding done at the same time. > Getting OS/2 code done always trumps everything else. Sure, you concentrate on getting wxWindows done. Just say when it's time for someone to try using autoconf with it... > Trust me, the best way to guarantee wide spread rejection of a consumer > targeted Open Source Windows-OS/2 application to make a user do "autoconf, > configure, make, make install". That's the quickest route to the Recycle > Box or Shredder. All I want as a Windows or OS/2 user is the classic > Setup.exe or Install.bat icon, nothing else will works if you want wide > spread acceptance. Far far too many Open Source zealots just don't get that. > They (the users) are NOT us (the developers). I don't think it's worth getting hung up about installation methods yet. We need to have something to install first. > -----Original Message----- > From: John Poltorak [mailto:jp at eyup.org] > Sent: Friday, July 26, 2002 8:34 AM > To: os2-unix at eyup.org > Subject: Re: wxWindows > > > You seem to have a narrow focus on VAC, > which is already out of support from IBM, so has no long term future. In > contrast OpenWatcom has only just been released and may provide stiff > competition to gcc, especially in open source projects over many years to > come. By exploring the benefits offered by autoconf+friends you may be > able to provide support for gcc and eventually Watcom without needing to > do anything special. > > Once there is a release of which works on OS/2, I'm sure people will see > how far things get using autoconf etc. Just think of it as a set of tools > for creating makefiles, libraries etc, automatically. The onus is on the > wxWindows developer to create the initial configure.in and makefile.in > files and everything else should fall into place. I guess we just need to > make sure that whatever is done for Linux is sufficiently cross platform > to work on OS/2 as well. > > > -- > John -- John **= Email 46 ==========================** Date: Fri, 26 Jul 2002 15:59:42 +0100 From: John Poltorak Subject: Re: Multiplatform LS On Fri, Jul 26, 2002 at 11:54:02AM -0300, Lord Spigol wrote: > Exactly what I need to compile? > > There is an Unix environment package for OS2 on hobbes called gnupack.wpi. > > Elsewhere there is the GCC on WPI also. > > I feel a bit losty. :) I've spent the last couple of weeks asking people to test a baseline development environment for building Unix apps on OS/2, which I have been trying to establish. Didn't you read any of the msgs? > []s > Rod > > On Fri, 26 Jul 2002 15:14:46 +0100, John Poltorak wrote: > > >make -f makefile.os2 -- John **= Email 47 ==========================** Date: Fri, 26 Jul 2002 16:14:43 +0100 From: John Poltorak Subject: Re: GLIBC On Fri, Jul 26, 2002 at 03:10:03PM +0200, Holger Veit wrote: > > > > All I'm looking for is a source for regex.c. There is one in GNU libc, but > > > it doesn't seem right to have to download a 16MB archive just to grab a > > > 250kB file... > [...] > > > > I'm not sure if this regex library is the one you're looking for > > because the BSD libc regex functions might differ from the GNU libc > > ones. > > There are multiple versions of standalone regex libraries on the net; > they usually fall in one of the following two categories. > Autoconf'd programs usually can deal with any of them. > > One style is: > re_comp() > re_exec() > the other is > regcomp() > regexec() > regerror() > regfree() > > You better search for the second one; this is the modern POSIX style. AFAIK that is the one in GLIBC. What I'm trying to find out is whether there is any way to grab a small subset of GLIBC just for the sake of building REGEX... I'm not sure how GLIBC is organised, it appears to be a package of many seperate apps, which, for all I know, may appear individually elsewhere, but I can't get a proper overview of the whole package. Is there a guide to GLIBC anywhere? > Holger > > -- > Please update your tables to my new e-mail address: > holger.veit$ais.fhg.de (replace the '$' with ' at ' -- spam-protection) -- John **= Email 48 ==========================** Date: Fri, 26 Jul 2002 16:38:31 +0100 From: John Poltorak Subject: Re: bootstrap.lst On Fri, Jul 26, 2002 at 07:53:48AM -0500, Maynard wrote: > John, > > I think that wget will maintain a connection over multiple files from > the same source, at least under some circumstances. If so, then if > bootstrap.lst were sorted ;-} ..... Would sorting it make it more efficient? > While looking at bootstrap.lst in this regard, I note that my latest > version received today, stamped 7-24-02 23:23, has a bad line in it: > > x04.zip > > without source I just checked here and it has (in *BASELINE*.LST) :- http://hobbes.nmsu.edu/pub/os2/util/archiver/uzs550x2.exe http://hobbes.nmsu.edu/pub/os2/util/archiver/zcr23x2.zip ftp://ftp.leo.org/pub/comp/os/os2/leo/gnu/emx+gcc/emxrt.zip http://hobbes.nmsu.edu/pub/os2/dev/unix/gettext-0_10_39-r2-bin.zip http://hobbes.nmsu.edu/pub/os2/util/file/gnufutil.zip http://hobbes.nmsu.edu/pub/os2/util/shell/gnusutil.zip http://hobbes.nmsu.edu/pub/os2/util/file/gnututil.zip http://www.math.ohio-state.edu/~ilya/software/os2/pdksh-5.2.14-bin-2.zip ftp://ftp.leo.org/pub/comp/os/os2/leo/gnu/shells/bash-os2-2.0-266.zip ftp://ftp.leo.org/pub/comp/os/os2/leo/gnu/systools/gnused.zip http://hobbes.nmsu.edu/pub/os2/dev/awk/gnuawk.zip http://hobbes.nmsu.edu/pub/os2/util/file/gnugrep.zip http://hobbes.nmsu.edu/pub/os2/dev/util/make-3_79_1-bin.zip http://hobbes.nmsu.edu/pub/os2/util/archiver/gtar258.zip http://hobbes.nmsu.edu/pub/os2/dev/misc/gnum4.zip http://hobbes.nmsu.edu/pub/os2/dev/util/gnupatch.zip http://hobbes.nmsu.edu/pub/os2/util/file/gnudiff.zip http://hobbes.nmsu.edu/pub/os2/util/disk/gnufind.zip ftp://ftp.leo.org/pub/comp/os/os2/leo/gnu/devtools/gnubison.zip ftp://ftp.leo.org/pub/comp/os/os2/leo/gnu/emx+gcc/emxfix04.zip http://www.math.ohio-state.edu/~ilya/software//os2/db_mt.zip It looks like the problem must have occurred in the course of the transfer... > Later, > > --Maynard -- John **= Email 49 ==========================** Date: Fri, 26 Jul 2002 16:46:21 -0500 From: Dave Webster Subject: RE: wxWindows Volunteer? I'm not going to do it, got better things to do, like finishing up the code for release? -----Original Message----- From: Henry Sobotka [mailto:sobotka at axess.com] Sent: Friday, July 26, 2002 4:23 PM To: os2-unix at eyup.org Subject: Re: wxWindows Dave Webster wrote: > > But so far wxWindows for OS/2 works only under VisualAge and EMX and both > have valid and seldom changing makefiles, so I see no need to futz with > autoconf at all, especially since it has no hope of working for VA ever or > Watcom any time soon. The Mozilla VA build has been using autoconf for several years now. If the developer has both VA and EMX installed, setting CC=icc will force VA mode. Otherwise, it automatically detects either gcc or icc and does the right thing. h~ **= Email 50 ==========================** Date: Fri, 26 Jul 2002 17:04:30 -0300 (ADT) From: "Lord Spigol" Subject: RE: Multiplatform LS Do you tried 4OS2? ftp://jpsoft.com/ On Fri, 26 Jul 2002 09:32:54 -0500, Dave Webster wrote: >By the same token is there a decent OS/2 cmd processor out there anywhere. >I've tried Hobbes and many other places. I really would like to get an OS/2 >command line window that I could adjust to something like a 72 x 132 with >full font and color selection like the one on Windows 2000 but I have yet to >find anything decent. **= Email 51 ==========================** Date: Fri, 26 Jul 2002 17:23:18 -0400 From: Henry Sobotka Subject: Re: wxWindows Dave Webster wrote: > > But so far wxWindows for OS/2 works only under VisualAge and EMX and both > have valid and seldom changing makefiles, so I see no need to futz with > autoconf at all, especially since it has no hope of working for VA ever or > Watcom any time soon. The Mozilla VA build has been using autoconf for several years now. If the developer has both VA and EMX installed, setting CC=icc will force VA mode. Otherwise, it automatically detects either gcc or icc and does the right thing. h~ **= Email 52 ==========================** Date: Fri, 26 Jul 2002 18:35:41 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap I notice that after the build, wget is not in /usr/bin - should it not be moved/copied from the build side of the fence? -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 53 ==========================** Date: Fri, 26 Jul 2002 18:58:08 +0200 (CEST) From: Stefan Neis Subject: RE: wxWindows On Fri, 26 Jul 2002, Dave Webster wrote: > When Open Source developers come to > understand, Open Source or not, they need to look like a traditional > commercial package, they'll get the next big increment of market > penetration. Until then, they won't. I think nobody is going to doubt those facts for the typical end-user application. For packages targetting developpers (especially cross-platform), things are slightly different. Personally, I tend to get irritated if I have to follow one particular installation scheme on Windows, another one on OS/2 and a third one on Linux for the same application/library. > I seriously doubt configure could generate the VA makefile mainly because it > is a very odd one due to the limitations of old VA's make with static > builds. So do I. The first thing is that all the required knowledge about a compilation environment has to have been worked into autoconf _first_. I seem to recall that some VA support was incorporated for building Mozilla/Netscape, but that might have been rather limited support. > Besides, VA users are much more familiar with VA's own MakeMake > than autoconf and autoconf is yet another learning curve (that goes for me, > too, and I already have a enough learning curves to live with and the older > I get the fewer I am willing to tolerate). I got familiar enough with autoconf to be able to modify the existing configure.in slightly so it works better with EMX (which basically is gcc, so autoconf/configure does basically know everything that's needed) but writing one from scratch would be way out of my league. Besides, since you have working makefiles, I don't see much point in trying to auto-generate them... The tmake stuff (yet another level of indirection) used by wxWindows to autogenerate the input files for autoconf/configure _and_ the Makefiles for the compilers without autoconf support (VC, Metrowerks, Borland and others) might be more interesting for that purpose - My impression of that system is that it might really be usable for VAC as well (it would essentially imply to strip the file lists out of the existing makefiles and replace them with an 'include' to an autogenerated file containing those file lists. And this autogenerated file essentially already exists for usage by EMX). Regards, Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 54 ==========================** Date: Fri, 26 Jul 2002 19:04:24 +0200 (CEST) From: Stefan Neis Subject: RE: wxWindows On Fri, 26 Jul 2002, Dave Webster wrote: > with in getting wxOS2 finished. I honestly do not have the time to > configure that AND get the last bit of coding done at the same time. > Getting OS/2 code done always trumps everything else. And rightly so! > The main point is that even though toolkits like wxWindows that offer cross > platform development, does NOT mean that the resultant applications have to > install the same way. That is actually a horrible idea. Applications > should install natively, the way users on each platform expect applications > to be delivered to them. I totally agree. > All I want as a Windows or OS/2 user is the classic > Setup.exe or Install.bat icon, nothing else will works if you want wide > spread acceptance. Personally, I also quite like the "unzip something.zip" way of installing things. Having to compile myself is _really_ inconvenient. In fact, I find myself using precompiled binaries even on Solaris and Linux almost exclusively. Things are of course slightly different for stuff where I get personally involved in _developping_. Regards, Stefan **= Email 55 ==========================** Date: Fri, 26 Jul 2002 19:14:53 +0200 (CEST) From: Stefan Neis Subject: Re: baseline perl harness results On Fri, 26 Jul 2002, Maynard wrote: > Failed 3/726 test scripts, 99.59% okay. 5/68656 subtests failed, 99.99% okay. That looks pretty good. > 63 tests and 563 subtests skipped. But I wonder why nobody is talking about those skipped tests. Don't those indicate a problem as well? I suppose the test suite is clever enough to not test for some of the more unix-like features on OS/2 like user-management and unix-like file permission, but does this account for all those skipped tests? Regards, Stefan -- Micro$oft is not an answer. It is a question. The answer is 'no'. **= Email 56 ==========================** Date: Fri, 26 Jul 2002 19:18:07 +0100 From: John Poltorak Subject: Re: baseline perl harness results On Fri, Jul 26, 2002 at 07:14:53PM +0200, Stefan Neis wrote: > On Fri, 26 Jul 2002, Maynard wrote: > > > Failed 3/726 test scripts, 99.59% okay. 5/68656 subtests failed, 99.99% okay. > > That looks pretty good. > > > 63 tests and 563 subtests skipped. > > But I wonder why nobody is talking about those skipped tests. Don't those > indicate a problem as well? Sure, there is a problem, but there isn't a handy summary that everyone can post showing which tests have been skipped. It is something we need to tackle, but I'd prefer to get the failing tests fixed first, if possible. > I suppose the test suite is clever enough to > not test for some of the more unix-like features on OS/2 like > user-management and unix-like file permission, but does this account for > all those skipped tests? If anyone knows how to get a summary of skipped tests, that would be useful. > Regards, > Stefan > -- > Micro$oft is not an answer. It is a question. The answer is 'no'. -- John **= Email 57 ==========================** Date: Fri, 26 Jul 2002 19:32:19 +0100 From: John Poltorak Subject: Re: UnixOS/2 bootstrap On Fri, Jul 26, 2002 at 06:35:41PM +0100, Dave Saville wrote: > I notice that after the build, wget is not in /usr/bin - should it > not be moved/copied from the build side of the fence? WGET is a bit of an oddball... The whole idea of the bootstrap falls down because the system depends on WGET already being available, and so it is subsequently ignore in putting all the apps in place. Ideally, I should use programs which are automatically available to the user in the first place so it would be better to use FTP for the initial file retrieval, since FTP is part of a standard OS/2 installation. Can anyone provide a simple FTP script? It is relatively easy to do, but it's many years since I last tried, and with the advent of WGET, it hasn't been necessary to use one. I guess I ought to add WGET to the baseline toolset too. > -- > Regards > > Dave Saville > Please note new email address dave.saville at ntlworld.com -- John **= Email 58 ==========================** Date: Fri, 26 Jul 2002 19:53:35 +0100 From: John Poltorak Subject: Re: Baseline EMX [emx_inst.cmd] On Fri, Jul 26, 2002 at 11:14:28AM -0500, Maynard wrote: > John, > > Part of the emx construction seeks to move /emx/lib/* to /usr/lib/ by > using 'mv' > > When the target directory /usr/lib already exists (even empty) the move > fails > mv: usr/lib: cannot overwrite directory > > and consequently the baseline is incorrect for future anticipations, > adding about a dozen failures to the harness test, perhaps most > notably; > ../lib/ExtUtils/t/Embed.t 9 9 100.00% 1-9 > > Since this move entails subdirectories of /emx/lib, a simple 'copy' > cannot be substituted; erasing the target first would be rude; 'xcopy' > could be used, followed by 'delete'. > > While solving this problem, lets [re]consider the use of moving /emx to > /usr > I will offer that emx is of enough significance (as is perl) to merit > the path of its own; but most particularly if it is likely to be > updated, as it would seem then to be most fortunate to ensure no > straggling relics. Downside of leaving /emx ? one additional target for > path and libpath ? I'm reluctant to move emx\bin to usr\bin because as it stands all the emx stuff is nicely self contained. I would like to remove some of it which I don't think is necessary but haven't identified which programs yet. As it stands the install works fine for a pristine environment, the complications only arise for a subsequent install. What I'd like to do is provide an uninstall, and/or option for leaving the already installed toolset in place. > -- > Maynard -- John **= Email 59 ==========================** Date: Fri, 26 Jul 2002 20:30:13 +0100 From: John Poltorak Subject: Re: baseline perl harness results On Fri, Jul 26, 2002 at 02:01:50PM -0500, Maynard wrote: > On Fri, 26 Jul 2002 19:18:07 +0100, John Poltorak wrote: > > >If anyone knows how to get a summary of skipped tests, that would be > >useful. > > grep for "skipping" in a log file for 'make test' will show which tests > were skipped in a form of > lib/Tie/File/t/21_win32..............skipping test on this platform > > looking at the harness log offers the same information I guess we need to decide whether to use the results from 'perl test' or 'perl harness'.... My preference would be 'perl test'. Anyone else?... -- John **= Email 60 ==========================** Date: Fri, 26 Jul 2002 20:55:26 +0100 From: John Poltorak Subject: Re: wxWindows On Fri, Jul 26, 2002 at 12:15:01PM -0500, Dave Webster wrote: > Just one more ??? on autoconf. If autoconf generates makefiles with POSIX > flags or requires "makes" to work with POSIX flags in makefiles, and POSIX > is a Unix specification what makes it so great as a "cross-platform" tool > for builds? POSIX is the Portable Operating System Interface specified by the IEEE. It has nothing to do with Unix AFAICT, although many flavours of Unix attempt to comply with POSIX. IMV it is a good thing for Open Source Software to comply with POSIX standards. See:- http://standards.ieee.org/regauth/posix/index.html -- John **= Email 61 ==========================** Date: Fri, 26 Jul 2002 21:18:49 +0100 (BST) From: "Dave Saville" Subject: Re: UnixOS/2 bootstrap On Fri, 26 Jul 2002 19:32:19 +0100, John Poltorak wrote: >Ideally, I should use programs which are automatically available to the >user in the first place so it would be better to use FTP for the initial >file retrieval, since FTP is part of a standard OS/2 installation. > >Can anyone provide a simple FTP script? In unix land one would do: ftp -nv <ftp.log open user anonymous my-email-address bin get /file file quit EOF And then check that the file had arrived by grepping ftp.log for "Transfer complete" as ftp does not set a return code on failure. OS/2 ftp does understand the command line switch, but the normal shell does not understand come-from ( the <<) . sh does of course - but then you have to get sh :-) However I just tried putting the commands into a file and pipeing that to ftp and that works: Here is the file called stuff: open bear user db password bin get LOGIN quit [E:\tmp]type stuff | ftp -nv IBM TCP/IP for OS/2 - FTP Client ver 09:55:17 on Apr 16 1999 Connected to bear. 220 bear FTP server (SunOS 5.7) ready. 331 Password required for db. 230 User db logged in. 200 Type set to I. 200 PORT command successful. 150 Binary data connection for LOGIN (192.168.0.2,3331) (609 bytes). 226 Binary Transfer complete. local: login remote: LOGIN 609 bytes received in 0.00 seconds (0 Kbytes/s) 221 Goodbye. So the script would be type | ftp -nv >ftp.log grep "Transfer complete" ftp.log >nul Check greps return code. -- Regards Dave Saville Please note new email address dave.saville at ntlworld.com **= Email 62 ==========================** Date: Fri, 26 Jul 2002 22:38:27 +0100 From: "Csaba" Subject: RE: wxWindows On 26 Jul 2002, at 9:04, Dave Webster wrote: > Actually, I am looking forward to WatcomC++ on OS/2. I thought it was > already available and I hadn't yet taken the plunge. Watcom is dead, long live OpenWatcom ! > I understand it comes > with complete support for STL??? > Not quite. The C++ dialect implemented by Watcom 11 is still quite old. It is possible to compile STLport with it, but without the iostream library. Ceci n'est pas un .signature **= Email 63 ==========================** Date: Fri, 26 Jul 2002 22:38:27 +0100 From: "Csaba" Subject: Re: wxWindows On 26 Jul 2002, at 15:49, John Poltorak wrote: > It _is_ already available. What I meant was it couldn't be used with > autoconf yet because it's make utility does not support POSIX flags which > is something that autoconf requires. Are you sure it's wmake the culprit ? I had the impression that the compiler itself does not conform to POSIX commandline standards (e.g. -o to set the output file) and autoconf cannot deal with *that* Ceci n'est pas un .signature **= Email 64 ==========================** Date: Fri, 26 Jul 2002 22:38:27 +0100 From: "Csaba" Subject: Re: UnixOS/2 bootstrap --Message-Boundary-3257 Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Mail message body On 26 Jul 2002, at 19:32, John Poltorak wrote: > On Fri, Jul 26, 2002 at 06:35:41PM +0100, Dave Saville wrote: > > I notice that after the build, wget is not in /usr/bin - should it > > not be moved/copied from the build side of the fence? > > WGET is a bit of an oddball... > > The whole idea of the bootstrap falls down because the system depends on > WGET already being available, and so it is subsequently ignore in putting > all the apps in place. > > Ideally, I should use programs which are automatically available to the > user in the first place so it would be better to use FTP for the initial > file retrieval, since FTP is part of a standard OS/2 installation. > I'm not sure FTP is able to work in a noninteractive way. > Can anyone provide a simple FTP script? > There is, however, RxFTP, a REXX extension which is able to do FTP. It's available from the OS/2 TCPIP package (at least for Warp4 Convenience Pack). I cobbled rexftp together (see attachment) Ceci n'est pas un .signature --Message-Boundary-3257 Content-type: text/plain; charset=US-ASCII Content-transfer-encoding: 7BIT Content-description: Text from file 'rexftp.cmd' /* REXX */ cd ftping BASE = "d:\tmp\ftping" rc = RxFuncAdd("FtpLoadFuncs","rxFtp","FtpLoadFuncs") rc = FtpLoadFuncs() rc = FtpSetUser('213.152.37.92','unixos2','') rc = FtpSetBinary("Binary") say '---------------------------------------------------------' call recurse '/pub/unixos2/build_system', BASE rc = FtpLogoff() exit /****************************************************/ recurse : procedure say '*** Entering directory 'arg(1)' for 'arg(2) rc = FtpChDir( arg(1) ) rc = FtpDir( '*', 'files.' ) /* This is what we get in the stem: drwxrwxrwx 1 postgres root 0 Jul 22 11:24 aux -rw-rw-rw- 1 postgres root 530 Jul 22 15:26 apps.lst */ do i = 1 to files.0 /*say files.i*/ parse var files.i perms one owner creator size month mday time filename if left( perms,1 ) = 'd' then do /* say filename' , a directory' say arg(1)||filename */ mkdir filename cd filename call recurse arg(1)||'/'||filename, arg(2) cd .. end else do say ' getting 'filename rc = FtpGet( filename, filename ) if rc \= 0 then say '!!! Failed : 'FTPERRNO end end say '*** Leaving directory 'arg(1) rc = FtpChDir('..') return --Message-Boundary-3257-- **= Email 65 ==========================** Date: Fri, 26 Jul 2002 22:38:27 +0100 From: "Csaba" Subject: Re: bootstrap.lst On 26 Jul 2002, at 7:53, Maynard wrote: > John, > > I think that wget will maintain a connection over multiple files from > the same source, at least under some circumstances. If so, then if > bootstrap.lst were sorted ;-} ..... And the other .lst files too benefit > > While looking at bootstrap.lst in this regard, I note that my latest > version received today, stamped 7-24-02 23:23, has a bad line in it: > > x04.zip > Yes, I had that too; this probably was emxfix04.zip some time before. Looks as if something was edited with a binary editor or something. Ceci n'est pas un .signature **= Email 66 ==========================** Date: Fri, 26 Jul 2002 22:38:27 +0100 From: "Csaba" Subject: Re: bootstrap.lst On 26 Jul 2002, at 11:36, Maynard wrote: > On Fri, 26 Jul 2002 16:38:31 +0100, John Poltorak wrote: > > >> I think that wget will maintain a connection over multiple files from > >> the same source, at least under some circumstances. If so, then if > >> bootstrap.lst were sorted ;-} ..... > > > >Would sorting it make it more efficient? > > wget has a "keep-alive" feature for persistent connections; however it > may only apply to http connections; I haven't figured it out completely > yet. Much of the baseline list is http anyway so it should work. I > recall seeing it work while actually getting the files; now that the > sources are not updated, the process appears to be making new > connections for each file > wget can reuse FTP logins as long as it doesn't have to login somewhere else. Also, if it has to download a bunch of files from the same FTP directory, it's smart enough not to change directories inbetween. Therefore, sorting the list would maximize the chances of similar URLs being next to each other, saving FTP logins (potentially the longest delays) and chdirs. Ceci n'est pas un .signature