Date: Sat, 22 Jan 2005 00:04:17 EST-10EDT,10,-1,0,7200,3,-1,0,7200,3600 Subject: [UnixOS2_Archive] No. 509 ************************************************** Friday 21 January 2005 Number 509 ************************************************** Subjects for today 1 Re: START from shell script : Dave Yeo" 2 Re: START from shell script : Dave Yeo" 3 Re: START from shell script : Dave Saville" 4 Re: START from shell script : Andreas Buening 5 Re: sed script to initialise variables : Steven Levine" 6 Re: START from shell script : John Poltorak 7 coreutils with gcc 3.3.5 b4 : Andy Willis 8 Re: coreutils with gcc 3.3.5 b4 : John Poltorak 9 Re: coreutils with gcc 3.3.5 b4 : Andy Willis 10 Re: coreutils with gcc 3.3.5 b4 : Steven Levine" 11 Re: coreutils with gcc 3.3.5 b4 : Andy Willis 12 Re: coreutils with gcc 3.3.5 b4 : Steve Wendt 13 Re: coreutils with gcc 3.3.5 b4 : Steven Levine" 14 Re: coreutils with gcc 3.3.5 b4 : Andy Willis 15 Re: coreutils with gcc 3.3.5 b4 : Dave Yeo" 16 Re: coreutils with gcc 3.3.5 b4 : Andy Willis 17 Re: coreutils with gcc 3.3.5 b4 : Andy Willis 18 Perl with GCC 3.3.5 B4 : lsunley at mb.sympatico.ca 19 Re: Perl with GCC 3.3.5 B4 : Henry Sobotka 20 Re: Perl with GCC 3.3.5 B4 : Dave Yeo" 21 Xfree86 with libc (was Re: Perl with GCC 3.3.5 B4) : Dave Yeo" 22 Re: Xfree86 with libc (was Re: Perl with GCC 3.3.5 B4) : Henry Sobotka 23 Re: Perl with GCC 3.3.5 B4 : Henry Sobotka 24 Re: START from shell script : Dave Saville" 25 Re: START from shell script : John Poltorak 26 Re: START from shell script : Dave Saville" 27 Re: coreutils with gcc 3.3.5 b4 : John Poltorak 28 Re: START from shell script : John Poltorak 29 Building PDKSH using latest gcc 3.3.5 : John Poltorak **= Email 1 ==========================** Date: Thu, 20 Jan 2005 07:15:12 -0800 From: "Dave Yeo" Subject: Re: START from shell script On Thu, 20 Jan 2005 11:30:20 +0000, John Poltorak wrote: > >Is it possible to start a new session from with a shell script? > >sh.exe doesn't understand the START command. Have you tried an ampersand? foo.exe &. I think the newest bash are the only port where it works right. Dave **= Email 2 ==========================** Date: Thu, 20 Jan 2005 07:54:43 -0800 From: "Dave Yeo" Subject: Re: START from shell script On Thu, 20 Jan 2005 07:15:12 -0800, Dave Yeo wrote: >On Thu, 20 Jan 2005 11:30:20 +0000, John Poltorak wrote: > >> >>Is it possible to start a new session from with a shell script? >> >>sh.exe doesn't understand the START command. > >Have you tried an ampersand? foo.exe &. I think the newest bash are the only port where it works right. >Dave Just tried this with ksh. Seems the two sessions are sharing the console and I can't kill them. I know it works with the latest bash. Dave **= Email 3 ==========================** Date: Thu, 20 Jan 2005 14:29:50 +0000 (GMT) From: "Dave Saville" Subject: Re: START from shell script On Thu, 20 Jan 2005 11:30:20 +0000, John Poltorak wrote: > >Is it possible to start a new session from with a shell script? > >sh.exe doesn't understand the START command. It's a *nix shell so: & Seems to work. -- Regards Dave Saville **= Email 4 ==========================** Date: Thu, 20 Jan 2005 20:59:07 +0100 From: Andreas Buening Subject: Re: START from shell script Dave Yeo wrote: > > On Thu, 20 Jan 2005 11:30:20 +0000, John Poltorak wrote: > > > > >Is it possible to start a new session from with a shell script? > > > >sh.exe doesn't understand the START command. > > Have you tried an ampersand? foo.exe &. I think the newest bash are the only port where it works right. > Dave If this doesn't work (with most shells) you may have to try cmd /c 'start the_command_you_want_to_start' Bye, Andreas **= Email 5 ==========================** Date: Thu, 20 Jan 2005 12:09:29 -0800 From: "Steven Levine" Subject: Re: sed script to initialise variables In <1CrGFt-09MDC40 at fwd08.sul.t-online.com>, on 01/19/05 at 02:52 PM, Stefan.Neis at t-online.de said: >I think sed is a bit more clever than that (and stops reading after the >second/third line, respectively). If not, then: T=`sed -n '2p;2q' Unless you're doing that dozens of times or for hundreds of variables, >rather keep it simple for the reader than for the computer... ;-) And more portable. Steven -- ---------------------------------------------------------------------- "Steven Levine" MR2/ICE 2.60b #10183 Warp4/FP15/14.100c_W4 www.scoug.com irc.fyrelizard.com #scoug (Wed 7pm PST) ---------------------------------------------------------------------- **= Email 6 ==========================** Date: Thu, 20 Jan 2005 20:51:39 +0000 From: John Poltorak Subject: Re: START from shell script On Thu, Jan 20, 2005 at 08:59:07PM +0100, Andreas Buening wrote: > Dave Yeo wrote: > > > > On Thu, 20 Jan 2005 11:30:20 +0000, John Poltorak wrote: > > > > > > > >Is it possible to start a new session from with a shell script? > > > > > >sh.exe doesn't understand the START command. > > > > Have you tried an ampersand? foo.exe &. I think the newest bash are the only port where it works right. > > Dave > > If this doesn't work (with most shells) you may have to try > cmd /c 'start the_command_you_want_to_start' Thanks for the suggestion. I can get that to work OK but I couldn't find a way of using '&' to do anything. > > Bye, > Andreas -- John **= Email 7 ==========================** Date: Thu, 20 Jan 2005 14:58:09 -0700 From: Andy Willis Subject: coreutils with gcc 3.3.5 b4 I have built the latest stable and alpha of coreutils (5.2.1 and 5.3.0 repectively). I ran configure and edited the makefile in src to add -Zomf to cflags and -lsocket to hostname ldadd. I also had to add for 5.3.0 #include in lib\chdir-long.c or you could define PATH_MAX to 260 which it is in syslimits.h. I also edited config.h to define drive letters and backslashes (can't tell a difference if I don't so far). Everything built except su.exe (no crypt module but it isn't useful anyway on OS/2). PWD in 5.2.1 works but the one in 5.3.0 just gives a rc of 5. RM works in both but not as expected. With the release I had that I got when setting up to build Mozilla will delete the contents of a folder foo if using rm -rf foo but the ones I am building give an error that foo is a directory. SORT from 5.2.1 gives a write configuration error of some sort but the one from 5.3.0 seems to work (at least based off using it to run configure for coreutils). This is what I have seen so far, don't know what alot of the stuff does so don't know how to test it. Andy **= Email 8 ==========================** Date: Thu, 20 Jan 2005 22:12:12 +0000 From: John Poltorak Subject: Re: coreutils with gcc 3.3.5 b4 On Thu, Jan 20, 2005 at 02:58:09PM -0700, Andy Willis wrote: > I have built the latest stable and alpha of coreutils (5.2.1 and 5.3.0 > repectively). This is excellent news! > I ran configure Ideally, you should run autoconf first and then you have a good chance of creating a more OS/2 friendly build of configure and shouldn't need to mess with makefiles. > and edited the makefile in src to add > -Zomf to cflags and -lsocket to hostname ldadd. I also had to add for 5.3.0 > #include in lib\chdir-long.c or you could define > PATH_MAX to 260 which it is in syslimits.h. I also edited config.h to > define drive letters and backslashes (can't tell a difference if I don't > so far). > Everything built except su.exe (no crypt module but it isn't useful > anyway on OS/2). > PWD in 5.2.1 works but the one in 5.3.0 just gives a rc of 5. RM works > in both but not as expected. With the release I had that I got when > setting up to build Mozilla will delete the contents of a folder foo if > using > rm -rf foo but the ones I am building give an error that foo is a > directory. SORT from 5.2.1 gives a write configuration error of some > sort but the one from 5.3.0 seems to work (at least based off using it > to run configure for coreutils). > This is what I have seen so far, don't know what alot of the stuff does > so don't know how to test it. It's good to know the package can almost build out of the box, but some of the individual files actually need porting so that they work correctly under OS/2. For instance cp (among others) should be able to handle EAs, and it's probably necessary to add support for wildcards - I don't think '*' in filenames works the same way under Unix BICBW... > Andy -- John **= Email 9 ==========================** Date: Thu, 20 Jan 2005 15:57:23 -0700 From: Andy Willis Subject: Re: coreutils with gcc 3.3.5 b4 John Poltorak wrote: >On Thu, Jan 20, 2005 at 02:58:09PM -0700, Andy Willis wrote: > > >>I have built the latest stable and alpha of coreutils (5.2.1 and 5.3.0 >>repectively). >> >> > >This is excellent news! > > > > >>I ran configure >> >> > >Ideally, you should run autoconf first and then you have a good chance of >creating a more OS/2 friendly build of configure and shouldn't need to >mess with makefiles. > > I had thought that configure did that. How do you force an autoconf? > > >>and edited the makefile in src to add >>-Zomf to cflags and -lsocket to hostname ldadd. I also had to add for 5.3.0 >> #include in lib\chdir-long.c or you could define >>PATH_MAX to 260 which it is in syslimits.h. I also edited config.h to >>define drive letters and backslashes (can't tell a difference if I don't >>so far). >>Everything built except su.exe (no crypt module but it isn't useful >>anyway on OS/2). >>PWD in 5.2.1 works but the one in 5.3.0 just gives a rc of 5. RM works >>in both but not as expected. With the release I had that I got when >>setting up to build Mozilla will delete the contents of a folder foo if >>using >>rm -rf foo but the ones I am building give an error that foo is a >>directory. SORT from 5.2.1 gives a write configuration error of some >>sort but the one from 5.3.0 seems to work (at least based off using it >>to run configure for coreutils). >>This is what I have seen so far, don't know what alot of the stuff does >>so don't know how to test it. >> >> > > >It's good to know the package can almost build out of the box, but some of >the individual files actually need porting so that they work correctly >under OS/2. For instance cp (among others) should be able to handle EAs, >and it's probably necessary to add support for wildcards - I don't think >'*' in filenames works the same way under Unix BICBW... > > > >>Andy >> >> > > > > I had guess I had noticed wild-cards too, now that you mention it, but didn't think of EA's. I am seeing a few other things as I try to use them to build Mozilla. Touch gets an access denied as did mv when trying to copy over an existing file. cp couldn't understand a path that the Mozilla build process through at it that contained a dot (eg. foo/foo/./foo/foo). I really don't know of any reason that the existing tools need replacing, I just wanted to see if I could build them. Andy **= Email 10 ==========================** Date: Thu, 20 Jan 2005 15:11:38 -0800 From: "Steven Levine" Subject: Re: coreutils with gcc 3.3.5 b4 In <41F03753.5070503 at comcast.net>, on 01/20/05 at 03:57 PM, Andy Willis said: >I am seeing a few other things as I try to use them to build Mozilla. >Touch gets an access denied as did mv when trying to copy over an >existing file. cp couldn't understand a path that the Mozilla build >process through at it that contained a dot (eg. foo/foo/./foo/foo). I >really don't know of any reason that the existing tools need replacing, >I just wanted to see if I could build them. If you get the urge, you might test md5sum. mdsum v2.0.11 can not check file using the md5 lists it builds. Try: md5sum * >tmp.md5 md5sum -c tmp.md5 Steven -- ---------------------------------------------------------------------- "Steven Levine" MR2/ICE 2.60b #10183 Warp4/FP15/14.100c_W4 www.scoug.com irc.fyrelizard.com #scoug (Wed 7pm PST) ---------------------------------------------------------------------- **= Email 11 ==========================** Date: Thu, 20 Jan 2005 16:27:16 -0700 From: Andy Willis Subject: Re: coreutils with gcc 3.3.5 b4 Steven Levine wrote: >In <41F03753.5070503 at comcast.net>, on 01/20/05 > at 03:57 PM, Andy Willis said: > > > >>I am seeing a few other things as I try to use them to build Mozilla. >>Touch gets an access denied as did mv when trying to copy over an >>existing file. cp couldn't understand a path that the Mozilla build >>process through at it that contained a dot (eg. foo/foo/./foo/foo). I >>really don't know of any reason that the existing tools need replacing, >>I just wanted to see if I could build them. >> >> > >If you get the urge, you might test md5sum. mdsum v2.0.11 can not check >file using the md5 lists it builds. Try: > > md5sum * >tmp.md5 > md5sum -c tmp.md5 > >Steven > > > > The * doesn't work. Is there an equivalent for unix? Andy **= Email 12 ==========================** Date: Thu, 20 Jan 2005 16:00:03 -0800 (PST) From: Steve Wendt Subject: Re: coreutils with gcc 3.3.5 b4 On Thu, 20 Jan 2005, Andy Willis wrote: >> md5sum * >tmp.md5 >> > The * doesn't work. Is there an equivalent for unix? On Unix, the shell expands * into all the filenames. You could try running it under sh/bash/pdksh/ash/whatever and see if you get a different result... **= Email 13 ==========================** Date: Thu, 20 Jan 2005 16:04:10 -0800 From: "Steven Levine" Subject: Re: coreutils with gcc 3.3.5 b4 In <41F03E54.8030408 at comcast.net>, on 01/20/05 at 04:27 PM, Andy Willis said: >>If you get the urge, you might test md5sum. mdsum v2.0.11 can not check >>file using the md5 lists it builds. Try: >> >> md5sum * >tmp.md5 >> md5sum -c tmp.md5 >> >The * doesn't work. Is there an equivalent for unix? Define doesn't work. Are sou waying the wildcard is not getting expanded? If so, you probably have a build problem. If you are getting permission denied messages for directories, this is expected. Steven -- ---------------------------------------------------------------------- "Steven Levine" MR2/ICE 2.60b #10183 Warp4/FP15/14.100c_W4 www.scoug.com irc.fyrelizard.com #scoug (Wed 7pm PST) ---------------------------------------------------------------------- **= Email 14 ==========================** Date: Thu, 20 Jan 2005 17:13:07 -0700 From: Andy Willis Subject: Re: coreutils with gcc 3.3.5 b4 Steven Levine wrote: >In <41F03753.5070503 at comcast.net>, on 01/20/05 > at 03:57 PM, Andy Willis said: > > > >>I am seeing a few other things as I try to use them to build Mozilla. >>Touch gets an access denied as did mv when trying to copy over an >>existing file. cp couldn't understand a path that the Mozilla build >>process through at it that contained a dot (eg. foo/foo/./foo/foo). I >>really don't know of any reason that the existing tools need replacing, >>I just wanted to see if I could build them. >> >> > >If you get the urge, you might test md5sum. mdsum v2.0.11 can not check >file using the md5 lists it builds. Try: > > md5sum * >tmp.md5 > md5sum -c tmp.md5 > >Steven > > > > Running it through ash worked. I then dropped out of ash and ran the check and they all showed ok except tmp.md5 which would have changed after reading itself. Andy **= Email 15 ==========================** Date: Thu, 20 Jan 2005 17:02:55 -0800 From: "Dave Yeo" Subject: Re: coreutils with gcc 3.3.5 b4 On Thu, 20 Jan 2005 14:58:09 -0700, Andy Willis wrote: > I ran configure and edited the makefile in src to add >-Zomf to cflags and -lsocket to hostname ldadd. I also had to add for 5.3.0 > #include in lib\chdir-long.c or you could define >PATH_MAX to 260 which it is in syslimits.h. I wonder where PATH_MAX is usually defined, I've had to add a few times now. Easier way to add -lsocket is to set LIBS=-lsocket before running configure. Seems that really hostname() should be in libc not libsocket since its in unistd.h. Did you get linker errors without -Zomf? Dave **= Email 16 ==========================** Date: Thu, 20 Jan 2005 18:43:51 -0700 From: Andy Willis Subject: Re: coreutils with gcc 3.3.5 b4 Steven Levine wrote: >In <41F03E54.8030408 at comcast.net>, on 01/20/05 > at 04:27 PM, Andy Willis said: > > > >>>If you get the urge, you might test md5sum. mdsum v2.0.11 can not check >>>file using the md5 lists it builds. Try: >>> >>> md5sum * >tmp.md5 >>> md5sum -c tmp.md5 >>> >>> >>> >>The * doesn't work. Is there an equivalent for unix? >> >> > >Define doesn't work. Are sou waying the wildcard is not getting expanded? >If so, you probably have a build problem. If you are getting permission >denied messages for directories, this is expected. > >Steven > > > Sorry, the * is being taken literal instead of as a wildcard. RC=0 [e:\cvs\work\mozilla\obj\dist\bin]md5sum * >tmp.md5 md5sum.exe: *: No such file or directory As Steve Wendt suggested though, it works as expected in ash. Andy **= Email 17 ==========================** Date: Thu, 20 Jan 2005 18:41:20 -0700 From: Andy Willis Subject: Re: coreutils with gcc 3.3.5 b4 Dave Yeo wrote: >On Thu, 20 Jan 2005 14:58:09 -0700, Andy Willis wrote: > > > >>I ran configure and edited the makefile in src to add >>-Zomf to cflags and -lsocket to hostname ldadd. I also had to add for 5.3.0 >>#include in lib\chdir-long.c or you could define >>PATH_MAX to 260 which it is in syslimits.h. >> >> > >I wonder where PATH_MAX is usually defined, I've had to add a few >times now. >Easier way to add -lsocket is to set LIBS=-lsocket before running configure. Seems that >really hostname() should be in libc not libsocket since its in unistd.h. >Did you get linker errors without -Zomf? >Dave > > > > > I almost think that syslimits.h should be called in limits.h but that's just what makes sense to me. I didn't know I needed to add -lsocket until after I had run configure and it was easy to add it. Yes I did get linker errors without -Zomf. Andy **= Email 18 ==========================** Date: Thu, 20 Jan 2005 19:57:45 -0500 From: lsunley at mb.sympatico.ca Subject: Perl with GCC 3.3.5 B4 Anyone doing a port with this? I cannot link with the 5.8.0 libs because GCC 3.3.5 insists on that leading "_" for all routines... Lorne -- ----------------------------------------------------------- lsunley at mb.sympatico.ca ----------------------------------------------------------- **= Email 19 ==========================** Date: Thu, 20 Jan 2005 23:04:24 -0500 From: Henry Sobotka Subject: Re: Perl with GCC 3.3.5 B4 lsunley at mb.sympatico.ca wrote: > Anyone doing a port with this? > > I cannot link with the 5.8.0 libs because GCC 3.3.5 insists on that > leading "_" for all routines... I was going to take a run at it over the holidays, but got worked dumped on me. Far easier and quicker than the port would be to do an EMX build with _foo=foo EXPORTS statements in the DEF file. But I have no idea if mixing C libs would have any side-effects, nor am I clear on gcc 3.x binary compatibility with stuff built with earlier versions. h~ **= Email 20 ==========================** Date: Thu, 20 Jan 2005 21:58:15 -0800 From: "Dave Yeo" Subject: Re: Perl with GCC 3.3.5 B4 On Thu, 20 Jan 2005 19:57:45 -0500, lsunley at mb.sympatico.ca wrote: >Anyone doing a port with this? > >I cannot link with the 5.8.0 libs because GCC 3.3.5 insists on that >leading "_" for all routines... What about just regenerating the OMF import library from the a.out ones? Perhaps by modifying \libc\usr\lib\makeomflibs.cmd. Knut did state some time back On Wed, 24 Sep 2003 02:38:39 +0200, Knut St. Osmundsen wrote: >If it's an import library you can usally just regenerate the OMF one >from the .a one. If you have a .def file you'll have to prefix all the >functions with '_' in most cases. Dave **= Email 21 ==========================** Date: Thu, 20 Jan 2005 22:02:14 -0800 From: "Dave Yeo" Subject: Xfree86 with libc (was Re: Perl with GCC 3.3.5 B4) On Thu, 20 Jan 2005 23:04:24 -0500, Henry Sobotka wrote: >I was going to take a run at it over the holidays, but got worked dumped > on me. Far easier and quicker than the port would be to do an EMX >build with _foo=foo EXPORTS statements in the DEF file. But I have no >idea if mixing C libs would have any side-effects, nor am I clear on gcc >3.x binary compatibility with stuff built with earlier versions. Just came across this problem trying to build Xfree86 v4.4 with Innotek_libc. Luckily all the DEF files are generated dynamically so it should just be a matter of hacking makedef.cmd. I wonder if it would better to go with _foo=foo or foo=_foo or just plain old foo. All DEF files use ordinals. Dave **= Email 22 ==========================** Date: Fri, 21 Jan 2005 01:30:41 -0500 From: Henry Sobotka Subject: Re: Xfree86 with libc (was Re: Perl with GCC 3.3.5 B4) Dave Yeo wrote: > > I wonder if it would better to go with _foo=foo or foo=_foo or just plain old foo. As I recall, the grammar is externalname=internalname so it has to be _foo=foo if the program you're linking from wants _foo. h~ **= Email 23 ==========================** Date: Fri, 21 Jan 2005 01:26:15 -0500 From: Henry Sobotka Subject: Re: Perl with GCC 3.3.5 B4 Dave Yeo wrote: > > What about just regenerating the OMF import library from the a.out ones? With Perl the OMF and a.out builds are completely different (no-fork and forkable). Can't mix them that way. h~ **= Email 24 ==========================** Date: Fri, 21 Jan 2005 07:54:52 +0000 (GMT) From: "Dave Saville" Subject: Re: START from shell script On Thu, 20 Jan 2005 20:51:39 +0000, John Poltorak wrote: > >Thanks for the suggestion. I can get that to work OK but I couldn't find >a way of using '&' to do anything. Clapped is part of my ClipView suite - it just waits until some text hits the clipboard. # # clipwait & [1] 17972 # pstat|grep CLIPWAIT 4635 4634 26 D:\BIN\CLIPWAIT.EXE 01 0200 FDFDA07C Block D:\BIN\CLIPWAIT.EXE 4635 26 D:\USR\LIB\EMX.DLL* # # # pstat|grep CLIPWAIT # First the shell runs clipwait - the [1] indicates the first detached process and the 17972 is the pid of that process. The PSTAT shows it. I then copied the PSTAT line to the clipboard and ran it again - clipwait has now terminated. However the shell won't terminate with an &ered process running. # which sh -rwxrwx--a 163844 Feb 15 2003 D:\BIN\sh.exe No idea what sh.exe is - it may be a renamed something else, but I have no idea how to find out. -- Regards Dave Saville **= Email 25 ==========================** Date: Fri, 21 Jan 2005 09:33:59 +0000 From: John Poltorak Subject: Re: START from shell script On Fri, Jan 21, 2005 at 07:54:52AM +0000, Dave Saville wrote: > However the shell won't terminate with an &ered process running. > > # which sh > -rwxrwx--a 163844 Feb 15 2003 D:\BIN\sh.exe > > No idea what sh.exe is - it may be a renamed something else, but I have no idea > how to find out. Have you installed UX2BS? That is the same filesize as I have. The datestamp would depend on when it was built. > -- > Regards > > Dave Saville > -- John **= Email 26 ==========================** Date: Fri, 21 Jan 2005 09:58:32 +0000 (GMT) From: "Dave Saville" Subject: Re: START from shell script On Fri, 21 Jan 2005 09:33:59 +0000, John Poltorak wrote: >On Fri, Jan 21, 2005 at 07:54:52AM +0000, Dave Saville wrote: > >> However the shell won't terminate with an &ered process running. >> >> # which sh >> -rwxrwx--a 163844 Feb 15 2003 D:\BIN\sh.exe >> >> No idea what sh.exe is - it may be a renamed something else, but I have no idea >> how to find out. > > >Have you installed UX2BS? That is the same filesize as I have. The >datestamp would depend on when it was built. Not recently :-) -- Regards Dave Saville **= Email 27 ==========================** Date: Fri, 21 Jan 2005 10:05:01 +0000 From: John Poltorak Subject: Re: coreutils with gcc 3.3.5 b4 On Thu, Jan 20, 2005 at 03:57:23PM -0700, Andy Willis wrote: > John Poltorak wrote: > > >On Thu, Jan 20, 2005 at 02:58:09PM -0700, Andy Willis wrote: > > > > > >>I have built the latest stable and alpha of coreutils (5.2.1 and 5.3.0 > >>repectively). > >> > >> > > > >This is excellent news! > > > > > > > > > >>I ran configure > >> > >> > > > >Ideally, you should run autoconf first and then you have a good chance of > >creating a more OS/2 friendly build of configure and shouldn't need to > >mess with makefiles. > > > > > I had thought that configure did that. How do you force an autoconf? Configure creates Makefiles for you and this is what people using Unix can do fairly safely, but it rarely works on OS/2. But if you recreate the configure script yourself using the latest version of autoconf for OS/2 you have a much better chance of ending up with the Makefile you need. Patching Makefiles is not the correct thing to do and should be avoided. Maybe you should take a look at UX2BS which provides a complete build environment including the most recent versions of Autoconf, Automake and Perl and fully automates the build process. Unfortunately it is based on gcc v2.8.1 at the moment, but I'd like to migrate it to v3.3.5+ before too long. > >It's good to know the package can almost build out of the box, but some of > >the individual files actually need porting so that they work correctly > >under OS/2. For instance cp (among others) should be able to handle EAs, > >and it's probably necessary to add support for wildcards - I don't think > >'*' in filenames works the same way under Unix BICBW... > > > > > > > >>Andy > >> > >> > > > > > > > > > I had guess I had noticed wild-cards too, now that you mention it, but > didn't think of EA's. > I am seeing a few other things as I try to use them to build Mozilla. > Touch gets an access denied as did mv when trying to copy over an > existing file. cp couldn't understand a path that the Mozilla build > process through at it that contained a dot (eg. foo/foo/./foo/foo). Getting the tools to build straight out of the box is important, but only half the battle. You also need to make them work properly, or as an OS/2 user would expect them to work, so some patches do need to be applied. I'd suggest you take a look at some of the patches made by KUR in OS/2 ports of the GNU *utils. Many of them are no longer needed because of the introduction of Autoconf and friends which build a platform dependent config.h at build time. This sorts out quite a number of problems, but you still need code changes like this, for example for MV:- *************** *** 197,202 **** --- 197,206 ---- /* chown turns off set[ug]id bits for non-root, so do the chmod last. */ + #ifdef OS2 + CopyEAs(source, dest); + #endif + /* Try to copy the old file's modtime and access time. */ { struct utimbuf tv; > I really don't know of any reason that the existing tools need > replacing, I just wanted to see if I could build them. It's good to be able to build the latest versions and have them available if required. The older versions do work in most cases, but sooner or later some maintainer will introduce a new feature in their app which demands v9.9 of gnu foo or whatever. > Andy -- John **= Email 28 ==========================** Date: Fri, 21 Jan 2005 10:51:42 +0000 From: John Poltorak Subject: Re: START from shell script On Fri, Jan 21, 2005 at 09:58:32AM +0000, Dave Saville wrote: > On Fri, 21 Jan 2005 09:33:59 +0000, John Poltorak wrote: > > >On Fri, Jan 21, 2005 at 07:54:52AM +0000, Dave Saville wrote: > > > >> However the shell won't terminate with an &ered process running. > >> > >> # which sh > >> -rwxrwx--a 163844 Feb 15 2003 D:\BIN\sh.exe > >> > >> No idea what sh.exe is - it may be a renamed something else, but I have no idea > >> how to find out. > > > > > >Have you installed UX2BS? That is the same filesize as I have. The > >datestamp would depend on when it was built. > > Not recently :-) Since the most recent released version of sh.exe by IlyaZ was 1/2/02, then it sounds like yours must have been built on your system. I must admit I didn't think it was two years since I made UX2BS available, although it may have been put there as part of an Perl build I created, where it was CP'ed to /bin as part of the build process. cp has the annoying habit of losing the original timestamp by default. Maybe grab IlyaZ's release and do a comp... > -- > Regards > > Dave Saville > -- John **= Email 29 ==========================** Date: Fri, 21 Jan 2005 11:35:05 +0000 From: John Poltorak Subject: Building PDKSH using latest gcc 3.3.5 Has anyone tried building PDKSH using the latest release of gcc 3.3.5? -- John