Date: Fri, 21 Feb 2003 02:35:18 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [Ux2bs_Archive] No. 84 ************************************************** Thursday 20 February 2003 Number 84 ************************************************** Subjects for today 1 Re: Sendmail 8.12.7 : John Poltorak 2 Re: Some help needed : John Poltorak 3 Re: Perl build on JFS : John Poltorak 4 Build Testing : Maynard" 5 Re: Perl build on JFS : Hakan" 6 Re: UPDATE_BASE.CMD success? : Hakan" 7 Still experimenting : brpms at earthlink.net 8 Re: Perl build on JFS : John Poltorak 9 Re: Perl build on JFS : Sebastian Wittmeier (ShadoW)" 10 Re: Perl build on JFS : Sebastian Wittmeier (ShadoW)" 11 Re: Perl build on JFS : Henry Sobotka 12 Re: Still experimenting : brpms at earthlink.net 13 Re: New Posix/2 build : brpms at earthlink.net 14 Re: Perl build on JFS : John Poltorak 15 ux2bs website : Maynard" 16 Re: Perl build on JFS : Sebastian Wittmeier (ShadoW)" 17 Re: UPDATE_BASE.CMD success? : John Poltorak 18 Re: Perl build on JFS : Stefan.Neis at t-online.de 19 Re: UPDATE_BASE.CMD success? : Stefan.Neis at t-online.de 20 Private archives ? : uunet 21 New Posix/2 build : Stefan.Neis at t-online.de 22 Re: Still experimenting : John Poltorak 23 Re: Private archives ? : Ted Sikora 24 Re: UPDATE_BASE.CMD success? : John Poltorak 25 Re: New Posix/2 build : Ted Sikora 26 Re: Still experimenting : Stefan.Neis at t-online.de 27 Re: Still experimenting : Stefan.Neis at t-online.de 28 Re: Private archives ? : John Poltorak **= Email 1 ==========================** Date: Fri, 21 Feb 2003 09:07:29 +0000 From: John Poltorak Subject: Re: Sendmail 8.12.7 On Thu, Feb 20, 2003 at 10:56:44PM -0500, Ted Sikora wrote: > John maybe we should bring this sendmail stuff to the os2-sendmail list. OK. Although it would be good if the Sendmail porter would actually join in too... BTW I just thought of adding '-x' to the 'exec sh ...' in sendmail\Build. It reveals:- + basename ../../sendmail/sendmail.0 + cp -p ../../sendmail/sendmail.0 sendmail.0.dist + [ -f helpfile ] + cd /unixos2/workdir/sendmail-8.12.7/obj.OS-2.2.i386/sendmail + cp -p ../../sendmail/helpfile . + rm -f /unixos2/workdir/sendmail-8.12.7/obj.OS-2.2.i386/sendmail/.settings7025 + echo divert(-1) + > /unixos2/workdir/sendmail-8.12.7/obj.OS-2.2.i386/sendmail/.settings7025 ../devtools/bin/Build[795]: internal error: alloc: freeing memory outside of block (corrupted?) > We are starting to clutter up and distract from UX2BS. We can add it > once it works... make that if! No, it *has* to work! There is a port of v8.12.3, so it can't be that hard to get this one built. > -- > Ted Sikora > tsikora at ntplx.net -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 2 ==========================** Date: Fri, 21 Feb 2003 10:13:29 +0000 From: John Poltorak Subject: Re: Some help needed On Thu, Feb 20, 2003 at 08:27:57PM -0700, brpms at earthlink.net wrote: > In <20030220213818.C83 at manninghammills.org>, on 02/20/03 > at 09:38 PM, John Poltorak said: > > > > Okay. I copied everything [I think] from e: to g:, and I replaced e: with > g: in ux2_env.cmd. > > I then did build cal. Very different result, short but no success. The > log file is attached. gcal-3.01.tar.gz already retrieved g:/unixos2/workdir g:/unixos2/workdir/gcal-3.01 build.sh[145]: autoconf: not found Using:- build.sh[145]: autoconf: not found ./configure loading site script g:/unixos2/lib/config.site loading cache ./config.cache Autoconf was not found so configure was not rebuilt. On occasion, the configure script, as included in the distribution will actually work on OS/2 as it is, but this is rare at the moment. To all intents and purposes you will need to run Autoconf as a first step to create an OS/2-friendly configure script. In copying files from one drive to another, you have overlooked something and now Autoconf doesn't work. It really is safest to start from scratch because then you will be dealing with a known environment. If you start manually copying things around, then no one else will know where anything is. > Can you point me to some pertinent tutorials? This is a steep learning cliff. Personal experience through trial and (lots of) error is the only way to learn how things work, along with feedback through other users. Bear in mind that this build framework is experimental and new things are being learned all the time. Incidentally, I did manage to build gcal, but it requires the additional step of running EMXBIND to turn it into an OS/2 executable. This is something which ought to be handled by the build system by way of the post-process scripts, but an error occurs whilst running Make. > *************************** > The same guy must have taken my attachment > *************************** > Paul Schwartz > > -- > ----------------------------------------------------------- > brpms at earthlink.net > ----------------------------------------------------------- -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 3 ==========================** Date: Fri, 21 Feb 2003 11:33:11 +0000 From: John Poltorak Subject: Re: Perl build on JFS On Thu, Feb 20, 2003 at 03:08:32PM -0500, Henry Sobotka wrote: > John Poltorak wrote: > > > > It seems as though this test:- > > > > lib/os2_ea.t > > > > always fails on JFS. > > No, the reason for the failure is that the name of the first EA created > by the test is '++'; names beginning with '.', '$', ' at ', '&' or '+' are > reserved for system use. Thus the call to write the EAs fails with > "invalid argument" and triggers the rest of the subtest failures. But the only people reporting this failure are those using JFS, so the test mustn't work correctly on JFS for some reason. > What I don't understand is why it doesn't show up among Paul's test > failures (along with rx_cmprt.t where it's a CRLF problem as someone on > this list pointed out a few months ago). the rx_cmprt.t test works if you have Object REXX installed. It seems that most of us don't use it. > basic.t is failing because PERL_SRC gets set to ../../.. when it should > be one directory deeper at that point. Not sure where to fix it in the > Byzantine build system, but once fed the files it's looking for, the > test passes. AIUI it there has been a fix made since v5.8.0 was released. > The two os2_process failures also appear due to a bug in the test, > though I'm less sure in this case and want to discuss it with Ilya. These are the tests which produce the greatest variation. Some people only fail one or two tests, usually 174 and 209, but others have a great many failures. Are these tests hardware related? > hostname.t is likely failing (don't get it here) for lack of a HOSTNAME > envar. Is that simply cured by setting HOSTNAME? If so, I'll set HOSTNAME to localhost in the standard build environment. Is that likely to cause any problems? > The stat.t failure also seems to be a local problem. It is also an intermittant problem. On my systems it occurs sometimes and sometimes it doesn't. I'm not aware of any changes which may have casued it to fail. What is it a test for anyway? > h~ -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 4 ==========================** Date: Fri, 21 Feb 2003 11:57:07 -0600 (CST) From: "Maynard" Subject: Build Testing This message reports successful logging and testing of packages built with update_base.cmd and provides a framework for use by any other batch command file or standalone call to 'build' The console (and logfile) displays after "build termcap": elapsed time: 22 secs end testing termcap installation ... ...success!! [apparently] and after failed "build byacc": Fri Feb 21 11:54:02 GMT 2003 elapsed time: 4 secs end testing byacc installation ... build.sh: X:/UnixOS2/scripts/test/byacc[1]: yacc: not found byacc failed It is my hope that John will like to incorporate these modifications, or your suggested improvements thereon, into his ux2bs distributions, so that further enhancements to this process can be enabled; modified build.sh and update_base.cmd; new file if_built.sh; new files in \unixos2\scripts\test\ My goal was to test each package built, and report success or failure, as part of the unattended batch processing; to be able to scan the report and quickly determine which packages completed successfully or failed. The primary component is incorporated into build.sh I inserted the following long lines into the end of the finish() function: echo testing $PKG installation ... if test -f $BLD_HOME/scripts/test/$PKG; then . $BLD_HOME/scripts/test/$PKG > $BLD_HOME/status/$PKG || { echo "$PKG failed" |tee -a $BLD_HOME/lib/build.err; exit 1 ; } else $PKG --version > $BLD_HOME/status/$PKG || { echo "$PKG failed" |tee -a $BLD_HOME/lib/build.err; exit 1 ; } fi echo ...success!! [apparently] exit 0 The default test is to run $PKG --version from shell; if that test is not appropriate for a package, for obvious instance any package with version in the name thereof, a test is defined in /unixos2/scripts/test/$package build.sh then exits back to build.cmd with proper exit code however, tee'ing the output in build.cmd causes build.cmd to always exit code 0 therefore, since build.cmd always returns success, something else must be used by other batchfiles such as update_base.cmd; so I copied the build.sh testing code to if_built.sh #!$UXRT/bin/sh # need one parameter for PKG # and need to test for BLD_HOME PKG=$1 echo testing $PKG installation ... if test -f $BLD_HOME/scripts/test/$PKG; then . $BLD_HOME/scripts/test/$PKG || { echo "$PKG failed"; exit 2 ; } else $PKG --version || { echo "$PKG failed" ; exit 2 ; } fi echo ...success!! [apparently] exit 0 which can now be called from update_base.cmd in order to control dependencies as follows: call build byacc sh if_built.sh byacc if errorlevel 1 goto skip_byacc call build flex :skip_byacc since flex depends upon byacc, if byacc fails, flex isn't attempted. I have not forced failures in order to discover other dependencies; but this system provides a method for managing them when they become known. later in update_base.cmd I grouped the unusual command set required for make: (call build make) && (setlocal & cd \unixos2\workdir\make-3.79.1 & make & copy make.exe \usr\bin & endlocal) After running this process through update_base.cmd, the file \unixos2\lib\build.err contains: byacc failed less failed and the directory \unixos2\status\ contains: 2-21-03 10:57 32 0 autoconf-2.13 2-21-03 10:57 0 0 byacc 2-21-03 11:04 32 0 ncurses 2-21-03 11:04 389 0 make-3.76.1 2-21-03 11:05 322 0 autoconf-2.50 2-21-03 11:05 289 0 automake-1.4-p5 2-21-03 11:06 389 0 make 2-21-03 11:07 301 0 autoconf 2-21-03 11:11 34 0 gettext 2-21-03 11:13 260 0 texinfo 2-21-03 11:14 277 0 automake 2-21-03 11:14 32 0 regex 2-21-03 11:15 30 0 termcap 2-21-03 11:15 0 0 less The zero length files indicate failed build test; the other file content is that output by the test, generally version information. Files requiring non-version testing, in \unixos2\scripts\test: autoconf-2.13 autoconf-2.50 automake-1.4-p5 byacc gettext make-3.76.1 make-3.79.1 ncurses regex termcap texinfo These files are intended to return success or failure by exitcode. For example, regex contains: test -e $uxrt/usr/include/regex.h && echo installed $uxrt/usr/include/regex.h Questions and suggestions hereby solicited. Thanks, -- Maynard _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 5 ==========================** Date: Fri, 21 Feb 2003 12:00:36 -0500 (EST) From: "Hakan" Subject: Re: Perl build on JFS I have also built perl on JFS with the following results: Failed 8/726 test scripts, 98.90% okay. 400/68652 subtests failed, 99.42% okay. Failed Test Stat Wstat Total Fail Failed List of Failed ------------------------------------------------------------------------ ------- ../ext/IO/lib/IO/t/io_sock.t 60 15360 20 19 95.00% 2-20 ../ext/Time/HiRes/HiRes.t 25 2 8.00% 7-8 ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 ../lib/Shell.t 1 256 4 1 25.00% 2 lib/os2_ea.t 21 8 38.10% 7-11 14-16 lib/os2_process.t 255 65280 227 184 81.06% 44-227 lib/os2_process_kid.t 227 184 81.06% 44-227 op/stat.t 73 1 1.37% 44 63 tests and 548 subtests skipped. I am disappointed and annoyed that we are getting different results, different errors etc. when building the very same application, perl. The different errors and test results are telling us that we are ending up with versions of perl which are different in some respects, however small, something that quite clearly can/will influence our ability to run various perl scripts and could result in various insidious errors. Hakan On Wed, 19 Feb 2003 10:04:41 +0000, John Poltorak wrote: > >I tried a Perl build on a JFS partition yesterday and got this result:- > > >Failed 7/726 test scripts, 99.04% okay. 24/68652 subtests failed, 99.97% 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/os2_ea.t 21 8 38.10% 7-11 14-16 >lib/os2_process.t 5 1280 227 5 2.20% 80 85 94 174 209 >lib/os2_process_kid.t 227 5 2.20% 80 85 94 174 209 >lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 >op/stat.t 73 1 1.37% 44 >63 tests and 557 subtests skipped. > > >It seems as though this test:- > >lib/os2_ea.t > >always fails on JFS. > >If you have had such a failure, could you report it to IlyaZ? >When I informed him about other people's results he seemed to dismiss it >and put any problems down to this build environment that was being used. > > >-- >John > > > > > >_______________________________________________ >UX2BS mailing list >UX2BS at powerusersbbs.net >http://powerusersbbs.net/mailman/listinfo/ux2bs > > _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 6 ==========================** Date: Fri, 21 Feb 2003 12:07:30 -0500 (EST) From: "Hakan" Subject: Re: UPDATE_BASE.CMD success? John, You are right, sorry about posting those log files. However, more importantly I assert that the build scripts are too forgiving and that we need to spend time developing build scripts which: * check for required components before any action is kicked off; * abort when an error is encountered and clearly state the nature of the error and how it should be remedied; * weed out warnings from errors (frankly there should be no warnings.) One should not have to be a Unix propeller-head to build the basic tools. Nor should one have to pour over voluminous build logs to separate the critical errors from the warnings. I looked quickly at the build scripts and saw that they are written in the DOS batch language, a much better approach would be to write them in Classic Rexx. John, I have already anticipated your reply -- "sounds good, why don't you do it :-)" -- so I will begin by looking at the perl build script over the weekend. Hakan On Sat, 15 Feb 2003 19:06:48 +0000, John Poltorak wrote: >On Sat, Feb 15, 2003 at 01:40:26PM -0500, Hakan wrote: >> I tried but ran into trouble -- it seems gettext may have been the only >> application built successfully. I am attaching my log files. >> >> Hakan > > >Hakan, > >Please don't send huge log files to the list. > > >What has happened in your case is that BYACC was not retrieved and most of >the subsequent builds failed as a result since yacc.exe is a prequisite >app for most of the other apps. > >What I need is to be able to exit early if WGET fails, but I'm not sure >the discussions on this earler were all that conclusive... > > > >> On Sat, 15 Feb 2003 12:59:55 +0000, John Poltorak wrote: >> >> > >> >Has anyone who managed to build Perl, had any success running >> >update_base.cmd? >> > >> >I'm wondering what other apps ought to be added to this initial program... >> > >> >So far, I've thought of adding termcap and less. >> > >> >sed, grep, awk and all the GNU utils will be added when they are known to >> >be working. >> > >> >PATCH appears to build correctly and I would add that but it doesn't >> >actually work properly... > > >-- >John > > > > > > > > > > > > > > > >_______________________________________________ >UX2BS mailing list >UX2BS at powerusersbbs.net >http://powerusersbbs.net/mailman/listinfo/ux2bs > > _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 7 ==========================** Date: Fri, 21 Feb 2003 12:35:49 -0700 From: brpms at earthlink.net Subject: Still experimenting I tried to start afresh without redownloading everything from the Web. I wiped out all the associated directories except for unixos2, adjusted ux2_bootstrap.cmd so that the same disk is used for bldrt and uxrt. Then I executed ux2_bootstrap.cmd. After it completed I did build perl. The result shown here was not as good as the first time with bldrt and uxrt on different disks [volumes]. What did I do wrong? Failed Test Stat Wstat Total Fail Failed List of Failed ---Failed 55/726 test scripts, 88.98% okay. 687/67628 subtests failed, 98.98% okay. ---------------------------------------------------------------------------- run/fresh_perl.t 97 72 74.23% 1-3 6 9 13-15 18-21 24-39 41-42 45-56 58-61 63-67 69-70 76-80 83-96 run/switches.t 19 17 89.47% 1-12 14-18 x2p/s2p.t 56 56 100.00% 1-56 (1 subtest UNEXPECTEDLY SUCCEEDED), 67 tests and 522 subtests skipped. [G:\unixos2\workdir\perl-5.8.0\t]cp perl.exe g:/usr/bin [G:\unixos2\workdir\perl-5.8.0\t]cp perl*.dll g:/usr/lib UXRT=g: TMP=g:\tmp PATH=g:\usr\bin;g:\emx\bin;g:\usr\local\bin;c:\os2; DPATH=c:\os2\system BLD_HOME=g:/unixos2 CFLAGS= ETC=g:\etc CONFIG_SITE= CPLUS_INCLUDE_PATH=g:/usr/include/cpp;g:/usr/include;g:/emx/include/cpp;g:/emx/include C_INCLUDE_PATH=g:/usr/include;g:/emx/include LIBRARY_PATH=g:/usr/lib;g:/emx/lib HOME=g:/home/root COMSPEC=c:\os2\cmd.exe REPOSITORY=g:/unixos2/archives/source BUILDLOGS=g:/unixos2/logs LDFLAGS= BLDRT=g: OS2_SHELL=c:\os2\cmd.exe TERM=ansi-color-3 OSRT=c: WORKDIR=g:/unixos2/workdir OS2PATH=g:\usr\bin;g:\emx\bin;g:\usr\local\bin;c:\os2; _emx_sig=00001d9b:00000000 The Operating System/2 Version is 4.50 Revision 14.086 Fri Feb 21 12:16:24 GMT 2003 elapsed time: 385 secs end -- ----------------------------------------------------------- brpms at earthlink.net ----------------------------------------------------------- _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 8 ==========================** Date: Fri, 21 Feb 2003 13:48:37 +0000 From: John Poltorak Subject: Re: Perl build on JFS On Fri, Feb 21, 2003 at 02:32:37PM +0100, Sebastian Wittmeier (ShadoW) wrote: > On Fri, 21 Feb 2003 11:33:11 +0000, John Poltorak wrote: > > >> What I don't understand is why it doesn't show up among Paul's test > >> failures (along with rx_cmprt.t where it's a CRLF problem as someone on > >> this list pointed out a few months ago). > > >the rx_cmprt.t test works if you have Object REXX installed. It seems that > >most of us don't use it. > > Object REXX accepts rexx code with Unix line terminators, classic rexx > does not. I could never determine where the rexx code that the test used was located. If it is in some file it should be easy enough to convert it before the test runs... > Sebastian -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 9 ==========================** Date: Fri, 21 Feb 2003 14:32:37 +0100 (CET) From: "Sebastian Wittmeier (ShadoW)" Subject: Re: Perl build on JFS On Fri, 21 Feb 2003 11:33:11 +0000, John Poltorak wrote: >> What I don't understand is why it doesn't show up among Paul's test >> failures (along with rx_cmprt.t where it's a CRLF problem as someone on >> this list pointed out a few months ago). >the rx_cmprt.t test works if you have Object REXX installed. It seems that >most of us don't use it. Object REXX accepts rexx code with Unix line terminators, classic rexx does not. Sebastian _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 10 ==========================** Date: Fri, 21 Feb 2003 14:56:49 +0100 (CET) From: "Sebastian Wittmeier (ShadoW)" Subject: Re: Perl build on JFS On Fri, 21 Feb 2003 13:48:37 +0000, John Poltorak wrote: >I could never determine where the rexx code that the test used was >located. If it is in some file it should be easy enough to convert it >before the test runs... That is easy. It is inside \unixos2\workdir\perl-5.8.0\t\lib\rx_cmprt.t itself. However, in the long term we need a better solution. Regina rexx exists on other operating systems, and we want to be compatible to ease porting Perl applications. Sebastian _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 11 ==========================** Date: Fri, 21 Feb 2003 15:23:09 -0500 From: Henry Sobotka Subject: Re: Perl build on JFS John Poltorak wrote: > > But the only people reporting this failure are those using JFS, so the > test mustn't work correctly on JFS for some reason. Perhaps because JFS protects the system against the use of the reserved characters, whereas HPFS and FAT rely on a sort of honor system. In any case, I don't think a test should use illegal characters and expect them to work. > > These are the tests which produce the greatest variation. Some people only > fail one or two tests, usually 174 and 209, but others have a great many > failures. Are these tests hardware related? No, most of it is testing opening and closing windows, sending messages to them and the like. I only got the two failures, which result from a window's x, y, flags, width, height not matching stored values after maximizing, minimizing etc. a few times. After several hours of playing around with it, and trying all the possibilities, I came to the conclusion that there was no way the two sets of values could match, possibly because of intervening window manipulation. I then discovered that moving the stored value down in the script to just before the final set of manipulations leading up to the test that fails eliminated the failure at 174, 209. as well as those in os2_process_kid.t. > Is that simply cured by setting HOSTNAME? Yes: [L:\Perl-5.8.0\t]perl ..\lib\Net\t\hostname.t 1..2 ok 1 ok 2 [L:\Perl-5.8.0\t]set HOSTNAME= [L:\Perl-5.8.0\t]perl ..\lib\Net\t\hostname.t 1..2 not ok 1 Use of uninitialized value in string eq at ..\lib\Net\t\hostname.t line 39 Use of uninitialized value in string eq at ..\lib\Net\t\hostname.t line 39 ok 2 > It is also an intermittant problem. On my systems it occurs sometimes and > sometimes it doesn't. I'm not aware of any changes which may have casued > it to fail. What is it a test for anyway? Basic file operations (open, close, seek, truncate, delete etc.). h~ _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 12 ==========================** Date: Fri, 21 Feb 2003 16:46:35 -0700 From: brpms at earthlink.net Subject: Re: Still experimenting In <18mKPN-0vTUn2C at fmrl11.sul.t-online.com>, on 02/21/03 at 10:08 PM, Stefan.Neis at t-online.de said: >That really looks like some serious failure in the build process occured. >Can you locate the line with "make install" in the log file and post the >20 or 30 lines before that so we can see what was the last command before >make install and what ent wrong with it? > Regards, > Stefan Yes I went to look at that and it was referring to stuff from my previous install and build. I'll have to clean up and start over. Thanks Paul Schwartz -- ----------------------------------------------------------- brpms at earthlink.net ----------------------------------------------------------- _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 13 ==========================** Date: Fri, 21 Feb 2003 17:19:18 -0700 From: brpms at earthlink.net Subject: Re: New Posix/2 build In <18mIme-0TQVV2C at fmrl05.sul.t-online.com>, on 02/21/03 at 08:24 PM, Stefan.Neis at t-online.de said: > Hi, >I uploaded a new build of Posix/2 to os2ports.com/incoming. >This one should fix the known issues of the previous one >(Note: new p2-gcc.exe, new include/sys/unistdx.h, new > cExt.{a,lib}). >Ted, is that you, who is maintaining the server? >I think it would be a good idea to delete the two older >versions (p2alpha*.zip and p2beta0_1_0.zip) from incoming, >now that a newer one is available. > Regards, > Stefan Don't find it there. Paul Schwartz -- ----------------------------------------------------------- brpms at earthlink.net ----------------------------------------------------------- _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 14 ==========================** Date: Fri, 21 Feb 2003 17:45:40 +0000 From: John Poltorak Subject: Re: Perl build on JFS On Fri, Feb 21, 2003 at 12:00:36PM -0500, Hakan wrote: > I have also built perl on JFS with the following results: > > Failed 8/726 test scripts, 98.90% okay. 400/68652 subtests failed, 99.42% okay. > Failed Test Stat Wstat Total Fail Failed List of Failed > ------------------------------------------------------------------------------- > ../ext/IO/lib/IO/t/io_sock.t 60 15360 20 19 95.00% 2-20 > ../ext/Time/HiRes/HiRes.t 25 2 8.00% 7-8 New to me. > ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 Bug in Perl Configure. > ../lib/Shell.t 1 256 4 1 25.00% 2 Dunno > lib/os2_ea.t 21 8 38.10% 7-11 14-16 This one seems to be common to everyone building on JFS. > lib/os2_process.t 255 65280 227 184 81.06% 44-227 > lib/os2_process_kid.t 227 184 81.06% 44-227 Everyone fails these two, but not to this extent. > op/stat.t 73 1 1.37% 44 A frequent failure, which has never been explained so far... > 63 tests and 548 subtests skipped. > > I am disappointed and annoyed that we are getting different results, > different errors etc. when building the very same application, perl. Yes, I, too, am a little frustrated that results are not exactly the same, but many are very similar, and we can account for most of the failures, which, thankfully are very small in number. > The different errors and test results are telling us that we are ending > up with versions of perl which are different in some respects, however > small, something that quite clearly can/will influence our ability to > run various perl scripts and could result in various insidious errors. I've done my best to try and create an isolated build shell, but there are some factors that it is difficult to take into account. Among these are the LIBPATH, the actual hardware being used and other things running concurrently. I would like to isolate things still further but don't see any way to do that. > Hakan > > On Wed, 19 Feb 2003 10:04:41 +0000, John Poltorak wrote: > > > > >I tried a Perl build on a JFS partition yesterday and got this result:- > > > > > >Failed 7/726 test scripts, 99.04% okay. 24/68652 subtests failed, 99.97% 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/os2_ea.t 21 8 38.10% 7-11 14-16 > >lib/os2_process.t 5 1280 227 5 2.20% 80 85 94 174 209 > >lib/os2_process_kid.t 227 5 2.20% 80 85 94 174 209 > >lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 > >op/stat.t 73 1 1.37% 44 > >63 tests and 557 subtests skipped. > > > > > >It seems as though this test:- > > > >lib/os2_ea.t > > > >always fails on JFS. > > > >If you have had such a failure, could you report it to IlyaZ? > >When I informed him about other people's results he seemed to dismiss it > >and put any problems down to this build environment that was being used. -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 15 ==========================** Date: Fri, 21 Feb 2003 18:31:15 -0600 (CST) From: "Maynard" Subject: ux2bs website Hi, I've updated my content at http://warped.mentabolism.org/UnixOS2/ux2bs.html which you're all welcome to reference, use, and take to wherever. I also welcome your inputs. Cheers/ux/2, -- Maynard _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 16 ==========================** Date: Fri, 21 Feb 2003 18:45:00 +0100 (CET) From: "Sebastian Wittmeier (ShadoW)" Subject: Re: Perl build on JFS On Fri, 21 Feb 2003 12:00:36 -0500 (EST), Hakan wrote: >The different errors and test results are telling us that we are ending >up with versions of perl which are different in some respects, We must check that first. Is the executable different or is it the system environment during testing? However, a different system environment influences the running of build scripts (like configure) and with it the executable. Sebastian _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 17 ==========================** Date: Fri, 21 Feb 2003 19:07:44 +0000 From: John Poltorak Subject: Re: UPDATE_BASE.CMD success? On Fri, Feb 21, 2003 at 12:07:30PM -0500, Hakan wrote: > However, more importantly I assert that the build scripts are too > forgiving and that we need to spend time developing build scripts > which: > * check for required components before any action is kicked off; > * abort when an error is encountered and clearly state the > nature of the error and how it should be remedied; It is often difficult to get meaningful error codes, especially out of something as complicated as Perl, where there are probably several failures along the way, but in spite of that a successful build of Perl occurs. > * weed out warnings from errors (frankly there should be no > warnings.) > > One should not have to be a Unix propeller-head to build the basic > tools. Of course not. All you should need to do is run:- build TOOL but we are not quite there yet. > Nor should one have to pour over voluminous build logs to > separate the critical errors from the warnings. These build logs provide useful information about where a build has gone wrong. Remember this build system is in its very early stages. The logs are not always so voluminous. Perl, definitely is not a typical app. > I looked quickly at the build scripts and saw that they are written in > the DOS batch language, You have the wrong impression. In many cases a build script is not required as the default build process which consists of:- autoconf configure make make install works perfectly well. Build scripts are only produced when the default process will not work. In some instances the build script only consists of a single command such as:- make -f os2\makefile.os2 gcc > a much better approach would be to write them > in Classic Rexx. John, I have already anticipated your reply -- > "sounds good, why don't you do it :-)" -- so I will begin by looking at > the perl build script over the weekend. Do you mean build_perl.cmd or Configure? If you mean Configure, I wish you the best of luck ;-)... > Hakan > -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 18 ==========================** Date: Fri, 21 Feb 2003 19:18:54 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl build on JFS Hi, > > I have also built perl on JFS with the following results: > > lib/os2_process.t 255 65280 227 184 81.06% 44-227 > > lib/os2_process_kid.t 227 184 81.06% 44-227 That reminds me very much of my first attempt of building it (which failed miserably). Too me, it looks like those tests are run on an incomplete build. Maybe searching for earlier errors in the build log would help (And modifying the 4 or 5 lines of the build script to only call make install and the test suite if the build actually succeeded would also be _very_ helpful). Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 19 ==========================** Date: Fri, 21 Feb 2003 20:20:30 +0100 From: Stefan.Neis at t-online.de Subject: Re: UPDATE_BASE.CMD success? Hi, > It is often difficult to get meaningful error codes, especially out of > something as complicated as Perl, where there are probably several > failures along the way, but in spite of that a successful build of Perl > occurs. Well, if "make" really does abort with an error (as opposed to internally handling the error and retrying the command (or issuing a slightly different one)), we can be quite confident that there is no point in trying make install and then running the test scripts. > In many cases a build script is not required as the default build process > which consists of:- > > autoconf > configure > make > make install > > works perfectly well. But even there it should be something like autoconf if errorlevel 1 goto end configure if errorlevel 1 goto end make if errorlevel 1 goto end make install :end Of course, a more refined version would be desirable, but for a starter, the above batch file could do... Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 20 ==========================** Date: Fri, 21 Feb 2003 20:23:02 +0000 From: uunet Subject: Private archives ? Hello ux2bs, Trying to get to the private archives I was greeted with: No file /ux2bs/ (/unixos2/home/mailman/archives/private/ux2bs/) Did I miss something ? -- Best regards, Csaba mailto:adwx88 at uk.uumail.com _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 21 ==========================** Date: Fri, 21 Feb 2003 20:24:31 +0100 From: Stefan.Neis at t-online.de Subject: New Posix/2 build Hi, I uploaded a new build of Posix/2 to os2ports.com/incoming. This one should fix the known issues of the previous one (Note: new p2-gcc.exe, new include/sys/unistdx.h, new cExt.{a,lib}). Ted, is that you, who is maintaining the server? I think it would be a good idea to delete the two older versions (p2alpha*.zip and p2beta0_1_0.zip) from incoming, now that a newer one is available. Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 22 ==========================** Date: Fri, 21 Feb 2003 21:04:01 +0000 From: John Poltorak Subject: Re: Still experimenting On Fri, Feb 21, 2003 at 12:35:49PM -0700, brpms at earthlink.net wrote: > I tried to start afresh without redownloading everything from the Web. If you want to save all your archives, you may want to set up an area completely seperate from unixos2. To do that you need to have the REPOSITORY variable set to something like v:/archives... If you are rebuilding an app you should delete its directory from WORKDIR or even delete that whole subdirectory tree. Better still start from scratch as you could easily have something leftover which could cause unforeseen results. > I wiped out all the associated directories except for unixos2, adjusted > ux2_bootstrap.cmd so that the same disk is used for bldrt and uxrt. Then I > executed ux2_bootstrap.cmd. After it completed I did build perl. ux2_bootstrap automatically builds Perl unless you edit build.lst. Did you start a perl build manually? > The > result shown here was not as good as the first time with bldrt and uxrt on > different disks [volumes]. What did I do wrong? Quite a lot obviously :-)... > Failed Test Stat Wstat Total Fail Failed List of > Failed ---Failed 55/726 test scripts, 88.98% okay. 687/67628 subtests > failed, 98.98% okay. > ---------------------------------------------------------------------------- > (1 subtest UNEXPECTEDLY SUCCEEDED), 67 tests and 522 subtests skipped. > > Fri Feb 21 12:16:24 GMT 2003 > elapsed time: 385 secs This is a bit of a give away. The build has done hardly anything at all It should take quite a lot longer - about an hour or more here. More like half an hour if you have a modern fast system. > > -- > ----------------------------------------------------------- > brpms at earthlink.net > ----------------------------------------------------------- > -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 23 ==========================** Date: Fri, 21 Feb 2003 21:17:46 -0500 From: Ted Sikora Subject: Re: Private archives ? If you guys really want archiving I can setup the maillist at OS2Ports? Will have to be Monday. That's the only time I will access to the MX records for sendmail. John Poltorak wrote: > On Fri, Feb 21, 2003 at 08:23:02PM +0000, uunet wrote: > >>Hello ux2bs, >>Trying to get to the private archives I was greeted with: >> >>No file /ux2bs/ (/unixos2/home/mailman/archives/private/ux2bs/) >> >>Did I miss something ? > > > Archiving is not yet functioning, but should be in the next version of > Mailman.. > > > >>-- >>Best regards, >> Csaba mailto:adwx88 at uk.uumail.com > > > Is that the real Csaba? Or an imposter? :-)... > -- -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 24 ==========================** Date: Fri, 21 Feb 2003 21:26:20 +0000 From: John Poltorak Subject: Re: UPDATE_BASE.CMD success? On Fri, Feb 21, 2003 at 08:20:30PM +0100, Stefan.Neis at t-online.de wrote: > Hi, > > > It is often difficult to get meaningful error codes, especially out of > > something as complicated as Perl, where there are probably several > > failures along the way, but in spite of that a successful build of Perl > > occurs. > > Well, if "make" really does abort with an error (as opposed to internally > handling the error and retrying the command (or issuing a slightly > different one)), we can be quite confident that there is no point in > trying make install and then running the test scripts. Agreed, but what about measuring how successful Configure has been? Presumably if a Makefile has been created, we can assume that Make should be run... > > In many cases a build script is not required as the default build process > > which consists of:- > > > > autoconf > > configure > > make > > make install > > > > works perfectly well. > > But even there it should be something like > > autoconf > if errorlevel 1 goto end How do you code this in a SHELL script? Maybe you could have a look at build.sh and suggest how it can be improved... > Regards, > Stefan -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 25 ==========================** Date: Fri, 21 Feb 2003 21:26:50 -0500 From: Ted Sikora Subject: Re: New Posix/2 build brpms at earthlink.net wrote: > In <18mIme-0TQVV2C at fmrl05.sul.t-online.com>, on 02/21/03 > at 08:24 PM, Stefan.Neis at t-online.de said: > > >> Hi, > > >>I uploaded a new build of Posix/2 to os2ports.com/incoming. >>This one should fix the known issues of the previous one >>(Note: new p2-gcc.exe, new include/sys/unistdx.h, new >>cExt.{a,lib}). > > >>Ted, is that you, who is maintaining the server? >>I think it would be a good idea to delete the two older >>versions (p2alpha*.zip and p2beta0_1_0.zip) from incoming, >>now that a newer one is available. Who? What? Where? OK I'll do it but I'm starting to get finger cramps like 'George Jetson' from all this button pressing. Think I should cleanup /incoming some of the stuff has been there for months. Anyone know a young blonde who would want to do file descriptions and archive the incoming directory? She can even sit in my lap. -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 26 ==========================** Date: Fri, 21 Feb 2003 22:08:36 +0100 From: Stefan.Neis at t-online.de Subject: Re: Still experimenting > executed ux2_bootstrap.cmd. After it completed I did build perl. The > result shown here was not as good as the first time with bldrt and uxrt on > different disks [volumes]. What did I do wrong? > > Failed Test Stat Wstat Total Fail Failed List of > Failed ---Failed 55/726 test scripts, 88.98% okay. 687/67628 subtests That really looks like some serious failure in the build process occured. Can you locate the line with "make install" in the log file and post the 20 or 30 lines before that so we can see what was the last command before make install and what ent wrong with it? Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 27 ==========================** Date: Fri, 21 Feb 2003 22:15:02 +0100 From: Stefan.Neis at t-online.de Subject: Re: Still experimenting Hi, > It should take quite a lot longer - about an hour or more here. More like > half an hour if you have a modern fast system. BTW, I just started building with current Posix/2. Looks quite good. Those hangs of "miniperl" I previously had during the build process _seem_ to have gone away. ;-) Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 28 ==========================** Date: Fri, 21 Feb 2003 22:31:26 +0000 From: John Poltorak Subject: Re: Private archives ? On Fri, Feb 21, 2003 at 08:23:02PM +0000, uunet wrote: > Hello ux2bs, > Trying to get to the private archives I was greeted with: > > No file /ux2bs/ (/unixos2/home/mailman/archives/private/ux2bs/) > > Did I miss something ? Archiving is not yet functioning, but should be in the next version of Mailman.. > -- > Best regards, > Csaba mailto:adwx88 at uk.uumail.com Is that the real Csaba? Or an imposter? :-)... -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs