Date: Sun, 2 Feb 2003 02:33:43 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [Ux2bs_Archive] No. 65 ************************************************** Saturday 01 February 2003 Number 65 ************************************************** Subjects for today 1 Re: Getting started with Passive FTP : IanM" 2 New Mailman Test : Ted Sikora 3 Perl - no go with p2-gcc : John Poltorak 4 Re: p2-gcc results : John Poltorak 5 Re: New Mailman Test : Ted Sikora 6 Re: p2-gcc results : Stefan.Neis at t-online.de 7 Re: gdbm 1.8.0 builds : Ted Sikora 8 Re: Perl - no go with p2-gcc : Ted Sikora 9 Re: p2-gcc results : John Poltorak 10 Perl : Ted Sikora 11 Re: gdbm 1.8.0 builds : John Poltorak 12 Re: p2-gcc results : Stefan.Neis at t-online.de 13 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 14 Re: Any more results? : Maynard" 15 Re: Any more results? : Maynard" 16 Re: Perl - no go with p2-gcc : John Poltorak 17 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 18 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 19 Re: Any more results? : Ted Sikora 20 Zope261b1 New and Improved! : Ted Sikora 21 Re: Perl - no go with p2-gcc : John Poltorak 22 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 23 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 24 Re: Any more results? : Bart van Leeuwen" 25 Re: Any more results? : John Poltorak 26 Re: Perl - no go with p2-gcc : Stefan.Neis at t-online.de 27 Build System changes : John Poltorak 28 Perl & Posix/2 : Stefan.Neis at t-online.de **= Email 1 ==========================** Date: Sun, 02 Feb 2003 02:20:47 +1100 (EDT) From: "IanM" Subject: Re: Getting started with Passive FTP Hi John >The reason I used FTP rather than WGET in the first place was that I only >wanted to use programs that were part of a standard OS/2 build and there >is no guarantee that people have WGET. IBM's FTP doesnt support passive unfortunetly. >Also I only suggest using WGET for retrieving the original script because >it is a shorthand way of telling people what to do. You could also >retrieve it using FTP or even just paste it in from Ted's web page. are, ok :-) >One thing that just struck me is, if you need to use Passive FTP, then >maybe RSYNC wouldn't work. Do you know that it actually will? RSYNC appears to have worked, I've got a swag of stuff downloaded (after I wiped everything for a nice, new, 'standards' based I just have to go back and find out why it couldnt run SED. Hmm, or maybe SED doesnt run because it didnt fully work ? I'll see what I can find out. Cheers IanM http://www.os2site.com/ "I am Homer of Borg, prepare to be assim...ooohhh DOOOOONUTS!" _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 2 ==========================** Date: Sun, 02 Feb 2003 11:00:45 -0500 From: Ted Sikora Subject: New Mailman Test Is it working? -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 3 ==========================** Date: Sun, 2 Feb 2003 11:10:11 +0000 From: John Poltorak Subject: Perl - no go with p2-gcc Perl fails to build when using p2-gcc. I left it to build overnight and have just been looking at the log, and found this error:- gcc -Zexe -Zomf -Zmt -Zcrtdll -Zstack 32000 -Zlinker /e:2 -o miniperl miniperlmain.obj perl.obj `echo malloc.obj gv.obj toke.obj perly.obj op.obj regcomp.obj dump.obj util.obj mg.obj reentr.obj hv.obj av.obj run.obj pp_hot.obj sv.obj pp.obj scope.obj pp_ctl.obj pp_sys.obj doop.obj doio.obj regexec.obj utf8.obj taint.obj deb.obj universal.obj xsutils.obj globals.obj perlio.obj perlapi.obj numeric.obj locale.obj pp_pack.obj pp_sort.obj os2.obj dl_os2.obj|sed -e 's/\bop\./opmini./g'` -lsocket -lm -lbsd -Zmap -Zlinker /map/PM:VIO LINK386 : warning L4051: cExt.lib : cannot find library I guess this means I need to run emxomf which I have just done like so:- emxomf -s -l -q cExt.a Hope this is OK. A cExt.lib got created but this msg was output:- emxomf: Input file `cExt.a(libuni.a)' is not an a.out file I don't know how serious this is. -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 4 ==========================** Date: Sun, 2 Feb 2003 11:30:11 +0000 From: John Poltorak Subject: Re: p2-gcc results On Sun, Feb 02, 2003 at 11:45:28AM +0100, Stefan.Neis at t-online.de wrote: > Hi, > > > Here is part of the diff of logs created whilst gettext was built using > > p2 and emx versions of gcc:- > > Summarizing your output, p2 adds str(n)casecmp, That in itself should bring a big sigh of relief. > > Are these results expected? > > Apart from the fact that "pathconf" which is included in p2 is not detected, > this looks pretty good. Will have a look at gettext's configure(.in) to see > what's going on ... > > > Should I include the posixised version of gcc in the build? > > Actually, if some software compiles and works I personally would still > tend to _not_ add Posix/2 into the compilation process ("If it's not > broken, don't fix it") when it comes to compile things for general usage. > On the other hand, using it and testing/distributing the results is the > only way to find remaining bugs. As I keep saying, this build system is work in progress. IMV, we need to get Posix/2 integrated and we can only do that by testing it rigorously. And building as many apps as possible seems as a good a way of testing as anything else. > My main problem is that I'm quite unsure about Posix/2's future status, > so it's difficult/impossible to decide whether it's worth the effort or > not... :-( If we want OS/2 to be able to build Open Source apps straight out of the box with a minimum number of hoops (which I do) then I think it is imperative to update the headers and libs that we use. The work already gone into Posix/2 is indispensible to making the build framework successful. In any case, what were the original goals for Posix/2? > Regards, > Stefan -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 5 ==========================** Date: Sun, 02 Feb 2003 11:31:27 -0500 From: Ted Sikora Subject: Re: New Mailman Test Reboot retest Ted Sikora wrote: > Is it working? > > -- > Ted Sikora > tsikora at ntplx.net > > > > _______________________________________________ > UX2BS mailing list > UX2BS at powerusersbbs.net > http://powerusersbbs.net/mailman/listinfo/ux2bs > > -- -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 6 ==========================** Date: Sun, 2 Feb 2003 11:45:28 +0100 From: Stefan.Neis at t-online.de Subject: Re: p2-gcc results Hi, > Here is part of the diff of logs created whilst gettext was built using > p2 and emx versions of gcc:- Summarizing your output, p2 adds str(n)casecmp, vfork, (un)setenv, tsearch, munmap and looses pathconf? > Are these results expected? Apart from the fact that "pathconf" which is included in p2 is not detected, this looks pretty good. Will have a look at gettext's configure(.in) to see what's going on ... > Should I include the posixised version of gcc in the build? Actually, if some software compiles and works I personally would still tend to _not_ add Posix/2 into the compilation process ("If it's not broken, don't fix it") when it comes to compile things for general usage. On the other hand, using it and testing/distributing the results is the only way to find remaining bugs. My main problem is that I'm quite unsure about Posix/2's future status, so it's difficult/impossible to decide whether it's worth the effort or not... :-( Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 7 ==========================** Date: Sun, 02 Feb 2003 12:01:39 -0500 From: Ted Sikora Subject: Re: gdbm 1.8.0 builds John Poltorak wrote: > On Sat, Feb 01, 2003 at 02:23:14PM -0500, Ted Sikora wrote: > >>To install the extra headers run make install-compats after make >>install. I been making the omf libs by adding the lib to the omflibs.cmd >>script then running it. What's command line for doing it or correct way? > > > Is it necessary to create omf libs? > > If so, is there any way this can be incorporated into the Makefile? > > Maybe this is something we could add into autoconf or automake... > Look at Kai Uwe Rommel's makefile in gnugdbm.zip He has modified it to make omf libs. -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 8 ==========================** Date: Sun, 02 Feb 2003 12:06:09 -0500 From: Ted Sikora Subject: Re: Perl - no go with p2-gcc Stefan.Neis at t-online.de wrote: > Hi, > > >>What about:- ? >> >>mv n*.h N*.h :-)... > > > Oops. :-) > Interesting idea. > But no, it doesn't work, it keeps spelling "netdb.h" > exactly the way it was given in the include statement. > > Regards, > Stefan > > P.S.: Was my recent tri-plicate due to hitting exactly the > moment where somebody was working on the list server > or is something wrong over here? > > I saved the db and installed it in the new mailman. I may have copied it over several times just to make sure I did not lose anything so it performed a pending send each time. -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 9 ==========================** Date: Sun, 2 Feb 2003 12:12:50 +0000 From: John Poltorak Subject: Re: p2-gcc results On Sun, Feb 02, 2003 at 12:54:42PM +0100, Stefan.Neis at t-online.de wrote: > Hi, > > > In any case, what were the original goals for Posix/2? > > I always thought it's a test field for ideas of the mytical > libemu. And I must admit I have no idea how much of the code > libemu will actually reuse and how much will be done from > scratch... I doubt whether the work on assembling the headers will be started from scratch. In any case many of them are BSD based so any new compilation would be very similar, at least as far as hose headers go. In any case EMX needs to be brought up to date so Posix/2 does give us something this irrespective of the future of libemu. > Regards, > Stefan -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 10 ==========================** Date: Sun, 02 Feb 2003 12:22:36 -0500 From: Ted Sikora Subject: Perl I installed Perl built with the ux2bs using the -Dprefix settings on the Mailman server and another Warp Server. Seems to be working fine. The DBI:: module installed perfect too. Best part no env setting except for path and libpath which is part of the ux2dist anyway. -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 11 ==========================** Date: Sun, 2 Feb 2003 12:50:46 +0000 From: John Poltorak Subject: Re: gdbm 1.8.0 builds On Sat, Feb 01, 2003 at 02:23:14PM -0500, Ted Sikora wrote: > > To install the extra headers run make install-compats after make > install. I been making the omf libs by adding the lib to the omflibs.cmd > script then running it. What's command line for doing it or correct way? Is it necessary to create omf libs? If so, is there any way this can be incorporated into the Makefile? Maybe this is something we could add into autoconf or automake... > -- > Ted Sikora > tsikora at ntplx.net -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 12 ==========================** Date: Sun, 2 Feb 2003 12:54:42 +0100 From: Stefan.Neis at t-online.de Subject: Re: p2-gcc results Hi, > In any case, what were the original goals for Posix/2? I always thought it's a test field for ideas of the mytical libemu. And I must admit I have no idea how much of the code libemu will actually reuse and how much will be done from scratch... Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 13 ==========================** Date: Sun, 2 Feb 2003 12:54:43 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, > emxomf -s -l -q cExt.a > > Hope this is OK. A cExt.lib got created but this msg was output:- > > emxomf: Input file `cExt.a(libuni.a)' is not an a.out file > > I don't know how serious this is. When I assembled the zip-File I thought that warning would be no problem, but meanwhile it looks a deadly error to me, at least my cExt.lib build that way apparently wasn't valid. Make a copy of cExt.a, say ccExt.a, then run ar d ccExt.a libuni.a uconv.a unikbd.a Run emxomf on the resulting ccExt.a and rename the generated lib file to cExt.lib Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 14 ==========================** Date: Sun, 02 Feb 2003 14:54:55 -0600 (CST) From: "Maynard" Subject: Re: Any more results? On Sun, 2 Feb 2003 19:54:35 +0000, John Poltorak wrote: >> ../lib/Net/t/hostname.t 2 1 50.00% 1 > >Problem with hostname resolution. > >I'd be interested in any suggested solutions for this. Maybe an entry in >HOSTS would suffice... This test will pass if there is an envar for DOMAIN. It's possible, and I keep forgetting to check, that the test will only be run if there is a HOSTS file in %etc% -- Maynard _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 15 ==========================** Date: Sun, 02 Feb 2003 15:22:18 -0600 (CST) From: "Maynard" Subject: Re: Any more results? On Sun, 02 Feb 2003 16:11:51 -0500, Ted Sikora wrote: >Weird that we all are getting different results. I would think that they >would be identical. The platform used must be part of the problem. I For this reason, John has been advocating, and developing, a common environment. In spite of his current success, there are still variables which cannot be overcome: format and size of hard drive volume; different Rexx flavors; different base OS and fix level and optional components (i.e. syslog, dns) existing network connections and protocols; different things which cause different tests to be run or not. We could develop and maintain a list of these notable differences for each report. os_base; fix_level; presume HPFS on %uxrt%; others as we identify them. Thanks, Ted, for your suggestion that two tests are WSeB particular. I've noted them at http://warped.mentabolism.org/UnixOS2/perl.html -- Maynard _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 16 ==========================** Date: Sun, 2 Feb 2003 15:43:02 +0000 From: John Poltorak Subject: Re: Perl - no go with p2-gcc On Sun, Feb 02, 2003 at 04:05:53PM +0100, Stefan.Neis at t-online.de wrote: > Hi, > > BTW, I found why Perl's configure doesn't detect "netdb.h" although > it should. In fact it's just good luck that it normally works. > _Any_ header not in the main "/usr/include" directory whose name is > starting with "n" won't be detected. > If the header is not in "/usr/include" (aka posix2/libext/include in > Posix/2, emx/include in standard EMX, /usr/include in your build environment > - and netdb.h is _not_ in posix2/libext/include as the EMX-Version can be > reused), a shell script is used to parse some C-Preprocessor output, which > looks like > # 1 "F:\\EMX\\INCLUDE\\netdb.h" 1 3 > Now, that stupid shell is _again_ doing it's nonsense on "\\" and the resulting > "\n" and tries to find "netdb.h" withing the two lines like > F:\EMX\INCLUDE > etdb.h > This, of course, fails - and it's more difficult to "fix" this one in the > compiler than making a gcc.exe which uses -lcExt. > > Three possible ways to fix it: > 1. Manually fix netdb.h in the Configure-generated config.h after running > configure. :-( > 2. "Fix" the test in the findhdr script generated by perl's configure. > 3. Copy posix2 headers over your /usr/include so all the headers are in > the main include directory again, so the buggy "advanced" test isn't used. What about:- ? mv n*.h N*.h :-)... > Regards, > Stefan > -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 17 ==========================** Date: Sun, 2 Feb 2003 16:05:53 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, BTW, I found why Perl's configure doesn't detect "netdb.h" although it should. In fact it's just good luck that it normally works. _Any_ header not in the main "/usr/include" directory whose name is starting with "n" won't be detected. If the header is not in "/usr/include" (aka posix2/libext/include in Posix/2, emx/include in standard EMX, /usr/include in your build environment - and netdb.h is _not_ in posix2/libext/include as the EMX-Version can be reused), a shell script is used to parse some C-Preprocessor output, which looks like # 1 "F:\\EMX\\INCLUDE\\netdb.h" 1 3 Now, that stupid shell is _again_ doing it's nonsense on "\\" and the resulting "\n" and tries to find "netdb.h" withing the two lines like F:\EMX\INCLUDE etdb.h This, of course, fails - and it's more difficult to "fix" this one in the compiler than making a gcc.exe which uses -lcExt. Three possible ways to fix it: 1. Manually fix netdb.h in the Configure-generated config.h after running configure. :-( 2. "Fix" the test in the findhdr script generated by perl's configure. 3. Copy posix2 headers over your /usr/include so all the headers are in the main include directory again, so the buggy "advanced" test isn't used. The first of course it not a real solution, just a quick hack, to be able to go on with compiling Perl. The second seems best to me, unfortunately, I have no idea how to do that or if it is possible at all without modifying the "sh" binary, so it looks like I'll be forced to take the third approach which implies using Posix/2 for everything or having to separate copies of "/usr/include". :-( Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 18 ==========================** Date: Sun, 2 Feb 2003 16:05:53 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, BTW, I found why Perl's configure doesn't detect "netdb.h" although it should. In fact it's just good luck that it normally works. _Any_ header not in the main "/usr/include" directory whose name is starting with "n" won't be detected. If the header is not in "/usr/include" (aka posix2/libext/include in Posix/2, emx/include in standard EMX, /usr/include in your build environment - and netdb.h is _not_ in posix2/libext/include as the EMX-Version can be reused), a shell script is used to parse some C-Preprocessor output, which looks like # 1 "F:\\EMX\\INCLUDE\\netdb.h" 1 3 Now, that stupid shell is _again_ doing it's nonsense on "\\" and the resulting "\n" and tries to find "netdb.h" withing the two lines like F:\EMX\INCLUDE etdb.h This, of course, fails - and it's more difficult to "fix" this one in the compiler than making a gcc.exe which uses -lcExt. Three possible ways to fix it: 1. Manually fix netdb.h in the Configure-generated config.h after running configure. :-( 2. "Fix" the test in the findhdr script generated by perl's configure. 3. Copy posix2 headers over your /usr/include so all the headers are in the main include directory again, so the buggy "advanced" test isn't used. The first of course it not a real solution, just a quick hack, to be able to go on with compiling Perl. The second seems best to me, unfortunately, I have no idea how to do that or if it is possible at all without modifying the "sh" binary, so it looks like I'll be forced to take the third approach which implies using Posix/2 for everything or having to separate copies of "/usr/include". :-( Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 19 ==========================** Date: Sun, 02 Feb 2003 16:11:51 -0500 From: Ted Sikora Subject: Re: Any more results? Weird that we all are getting different results. I would think that they would be identical. The platform used must be part of the problem. I rebuilt perl at least a hundred times with who knows how many changes. The 2 that remain and are distinctly different are the socketpair() and shell.t fails. It has to be a OS/2 WSeB issue. These seem to be the common problems. basic.t os2_process.t os2_process_kid.t rx_cmprt.t Here's mine: Failed Test Stat Wstat Total Fail Failed List of Failed ------------------------------------------------------------------------------- ../ext/Socket/socketpair.t 4 1024 45 4 8.89% 16-19 ../lib/ExtUtils/t/basic.t 1 256 17 1 5.88% 14 ../lib/Shell.t 1 256 4 1 25.00% 2 lib/os2_process.t 2 512 227 2 0.88% 174 209 lib/os2_process_kid.t 227 2 0.88% 174 209 lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 58 tests and 564 subtests skipped. Failed 6/726 test scripts, 99.17% okay. 13/68693 subtests failed, 99.98% okay. Bart van Leeuwen wrote: > Failed 7/726 test scripts, 99.04% okay. 20/68650 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 3 768 227 3 1.32% 90 174 209 > lib/os2_process_kid.t 227 3 1.32% 90 174 209 > lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 > op/stat.t 73 1 1.37% 44 > 64 tests and 557 subtests skipped. > -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 20 ==========================** Date: Sun, 02 Feb 2003 16:30:39 -0500 From: Ted Sikora Subject: Zope261b1 New and Improved! New build made with the UX2BS(modified). Some env cleanups and better xml support. Runs as smooth as a baby's bottom. ftp://os2ports.com/incoming/Zope261b1-os2emx-030202.zip -- Ted Sikora tsikora at ntplx.net _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 21 ==========================** Date: Sun, 2 Feb 2003 16:40:08 +0000 From: John Poltorak Subject: Re: Perl - no go with p2-gcc On Sun, Feb 02, 2003 at 05:26:43PM +0100, Stefan.Neis at t-online.de wrote: > Hi, > > > What about:- ? > > > > mv n*.h N*.h :-)... > > Oops. :-) > Interesting idea. > But no, it doesn't work, it keeps spelling "netdb.h" > exactly the way it was given in the include statement. Where abouts exactly is this test? Maybe Within the ux2bs environment, we could change all occurences of '\\' to '/' if they refer to directory seperators > Regards, > Stefan > > P.S.: Was my recent tri-plicate due to hitting exactly the > moment where somebody was working on the list server > or is something wrong over here? -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 22 ==========================** Date: Sun, 2 Feb 2003 17:26:43 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, > What about:- ? > > mv n*.h N*.h :-)... Oops. :-) Interesting idea. But no, it doesn't work, it keeps spelling "netdb.h" exactly the way it was given in the include statement. Regards, Stefan P.S.: Was my recent tri-plicate due to hitting exactly the moment where somebody was working on the list server or is something wrong over here? _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 23 ==========================** Date: Sun, 2 Feb 2003 18:51:16 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, > > But no, it doesn't work, it keeps spelling "netdb.h" > > exactly the way it was given in the include statement. > > Where abouts exactly is this test? If you look at perl's Configure, it writes a file "findhdr" at lines 5020 to 5054. This findhdr is a shell-script that is put into the "uu" subdirectory - if you look directly at that, it's slightly more readable than the configure lines that generate it. findhdr now tests for $usrincdir/$wanted, which in the case of Posix/2 and with wanted=netdb.h expands to /posix2/libext/include/netdb.h, which does not exist (but /emx/include/netdb.h or /usr/include/netdb.h does exist, so it _should_ still work). So findhdr continues with the second part, i.e. gcc -E ... | g:/usr/bin/grep g:/usr/bin/grep "^[ ]*#.*$wanted" | while read cline; do I broke that into pieces and found that "gcc -E " gives output containing "some_path\\netdb.h" which is passed unmodified through grep, but the "while read cline" apparently interprets \\ and \n and "cline" contains the "wrongly" replaced characters and the test fails ... :-( > Maybe Within the ux2bs environment, > we could change all occurences of '\\' > to '/' if they refer to directory seperators Those '\\' are generated by "gcc -E" ... OK, thanks for asking good questions, I found something. :-) If I pass the grep output through sed (replacing "\\" by "/") prior to passing it to "read", everything works fine... Now there's a nightmare of escape characters involved into getting this into Configure correctly... Here's a suitable diff for Perl's configure: --- perl-5.8.0\Configure Thu Jul 18 22:55:14 2002 +++ ./Configure Sun Feb 2 18:48:52 2003 at at -5032,6 +5032,7 at at echo "#include <\$wanted>" > foo\$\$.c $cppstdin $cppminus $cppflags < foo\$\$.c 2>/dev/null | \ $cppfilter $grep "^[ ]*#.*\$wanted" | \ +sed "s/\\\\\\\\\\\\\\\\/\//g"| \ while read cline; do name=\`echo \$cline | $awk "\$awkprg" | $tr -d '"'\` case "\$name" in Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 24 ==========================** Date: Sun, 2 Feb 2003 19:02:25 +0100 From: "Bart van Leeuwen" Subject: Re: Any more results? Failed 7/726 test scripts, 99.04% okay. 20/68650 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 3 768 227 3 1.32% 90 174 209 lib/os2_process_kid.t 227 3 1.32% 90 174 209 lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 op/stat.t 73 1 1.37% 44 64 tests and 557 subtests skipped. _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 25 ==========================** Date: Sun, 2 Feb 2003 19:54:35 +0000 From: John Poltorak Subject: Re: Any more results? On Sun, Feb 02, 2003 at 07:02:25PM +0100, Bart van Leeuwen wrote: We're seeing the same old failures, so it looks as though we almost have standard environment... > Failed 7/726 test scripts, 99.04% okay. 20/68650 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 Perl build bug... > ../lib/Net/t/hostname.t 2 1 50.00% 1 Problem with hostname resolution. I'd be interested in any suggested solutions for this. Maybe an entry in HOSTS would suffice... > lib/os2_ea.t 21 8 38.10% 7-11 14-16 This seems to occur from time to time. It must be realted to extended attributes and I can only guess that the partition being used does not support them. Can anyone else think of a reason for this to fail? > lib/os2_process.t 3 768 227 3 1.32% 90 174 209 > lib/os2_process_kid.t 227 3 1.32% 90 174 209 Usual failures. > lib/rx_cmprt.t 255 65280 18 3 16.67% 16-18 Missing Object REXX. There is a patch for this, but it hasn't been included yet. > op/stat.t 73 1 1.37% 44 This is one I get intermittantly but haven't found any pattern to it. > 64 tests and 557 subtests skipped. -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 26 ==========================** Date: Sun, 2 Feb 2003 20:22:38 +0100 From: Stefan.Neis at t-online.de Subject: Re: Perl - no go with p2-gcc Hi, One more patch for compiling perl with Posix/2: In hints/os2.sh, we search for a line like "char * _ptr" in stdio.h with an arbitrary _positive_ (i.e. non-zero) number of spaces both before and after "*". However, Posix/2's stdio.h has "char* _ptr", which is completely legal but fails the test. The attached patch to hints/os2.sh fixed that problem. Also, if you aren't using the (highly experimental) "p2-gcc", this patch add -lcExt at the right place. (In fact, that's probably even better than relying on p2-gcc which would add -lcExt _after_ the other libraries in this statement, with this patch, it's placed _before_ them!) Regards, Stefan ---------- patch: ------------------ --- os2.sh~ Sat Jun 1 19:03:12 2002 +++ os2.sh Sun Feb 2 20:12:46 2003 at at -213,7 +213,7 at at # [Maybe we should just remove c from $libswanted ?] # Test would pick up wrong rand, so we hardwire the value for random() -libs='-lsocket -lm -lbsd' +libs='-lcExt -lsocket -lm -lbsd' randbits=31 archobjs="os2$obj_ext dl_os2$obj_ext" at at -229,7 +229,7 at at # for speedup: (some patches to ungetc are also needed): # Note that without this guy tests 8 and 10 of io/tell.t fail, with it 11 fails -stdstdunder=`echo "#include " | cpp | egrep -c "char +\* +_ptr"` +stdstdunder=`echo "#include " | cpp | egrep -c "char *\* *_ptr"` d_stdstdio='define' d_stdiobase='define' d_stdio_ptr_lval='define' _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 27 ==========================** Date: Sun, 2 Feb 2003 20:36:40 +0000 From: John Poltorak Subject: Build System changes I have made a couple of changes to the build system, principally to facilitate the incorporation of Posix/2. Now the Posix/2 headers and lib get installed in \usr\include and \usr\lib and then new posixised version of gcc.exe gets installed in \usr\bin. The old one remains in \emx\bin. Due to ongoing problems getting Perl built in a Posix/2 environment, build_perl.cmd has been slightly modified to protect it from the new environment so it builds with the old headers and libs and the original gcc, but everything else should build with the new environment. One other enhancement is that %repository% is only set if it does not already exist. This means that if you want to use something like a network repository where have already downloaded many archives, you need to set it before running ux2_bootstrap. Just make sure that you have the correct directory structure. Currently, I'm using a TVFS drive and have set repository=v:/archives which looks like this:- V:\archives\baseline V:\archives\baseline\emx V:\archives\source This is what I run on the server:- tvlink t:\archives d:\unixos2\archives I have an ALIAS called TVFS which is set up as T:\ and then from anywhere on the network you should be able to run:- net use v: tvfs which gives you access to v: If anyone has any problems with any of the changes, please let me know. -- John _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs **= Email 28 ==========================** Date: Sun, 2 Feb 2003 21:13:16 +0100 From: Stefan.Neis at t-online.de Subject: Perl & Posix/2 Hi, After working around the various compile time problems, I now get to the linking stage, which fails with a most interesting error, since perl code provides replacements for some of the functions not existing in EMX but existing in Posix/2 (multiple defined symbol linker errors). This concerns some functions related to parsing /etc/group file (getgrent, setgrent, endgrent) and set/getpriority. It's probably no problem at all to throw Perl's stubs for the first three functions away, but just throwing get/setpriority away might have bad results. I didn't compare the code line by line, but Perl's and Posix/2's mapping of the standard unix priorities into the four big priority classes of OS/2 (time critical, foreground, ...) might be different, so switching from Perl's implementation to Posix/2's implementation _could_ cause a surprising change in behaviour... Regards, Stefan _______________________________________________ UX2BS mailing list UX2BS at powerusersbbs.net http://powerusersbbs.net/mailman/listinfo/ux2bs