Re: GCCCMS for zVM
--- Rick Troth <[EMAIL PROTECTED]> wrote: > I wonder: could you get it to call LE libraries > instead of OS libs? > > -- R; > Rick, I know its a long time since you asked the above, but looking at the original Dave Pitts compiler, I note he did originally provide a port to LE on MVS. I'll try and have a look at that and see how easy/hard it would be to build a VM version. Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
Sounds great. I look forward to downloading/uploading the latest level, then I can try learning C by actually trying to port some code. /Tom Kern --- Dave Wade <[EMAIL PROTECTED]> wrote: > Glad its been od some use. Just to say that whilst I > have worked hard, a lot of work has goine on behind > the scenes. Dave Pitts who started this has helped fix > some of the machine definitaions which were wrong for > unsigned half words. Phil Roberts has been helping > with math.h & of course Paul Edwards has moved the > support from GCC322 to GCC323. > > Well the new version should be much cleaner in ESA. I > have :- > > 1) Changed the startup code to issue FILEDEFS for the > console if SYSPRINT and SYSIN are not defined, so you > can now run modules with no wrapper EXEC. > > 2) Re-assembled everything with proper AMODE/RMODE > settings, and tweaked the support code to issue > GETMAIN/FREEMAIN so it will get 31 bit storage. > > 3) Cleaned up the assembler support routine so it uses > CMSCALL to issue CMS commands. The routine is still > called as _SVC202_ from C defined as follows:- > > _SVC202_(PARMS,CODE,ERROR) > > WHERE :- > PARMS IS A POINTER TO AN SVC202 PARAMETER LIST > CODE IS A CODE TO SAY OF &CONTROL IS ON OR OFF > ERROR is the RC from the CMSCALL. > > In addition there is some support for floating point. > > a. The s/f/printf, now has soem basic support for > doubles(). > > b. strtod() works for numbers with a fractional part. > > c. There is a preliminary "alpha" release of support > for "math.c" > > Sorting out the runtime to use natice CMS I/O is high > on my ToDo list, along with cleaning up the > implementation of math.h. I think math.h is next as > "Cody and Waite" arrived on an inter-library loan so I > need to finish it before I have to send it back:-) > > Hope this is all OK > > Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
Tom, Glad its been od some use. Just to say that whilst I have worked hard, a lot of work has goine on behind the scenes. Dave Pitts who started this has helped fix some of the machine definitaions which were wrong for unsigned half words. Phil Roberts has been helping with math.h & of course Paul Edwards has moved the support from GCC322 to GCC323. Well the new version should be much cleaner in ESA. I have :- 1) Changed the startup code to issue FILEDEFS for the console if SYSPRINT and SYSIN are not defined, so you can now run modules with no wrapper EXEC. 2) Re-assembled everything with proper AMODE/RMODE settings, and tweaked the support code to issue GETMAIN/FREEMAIN so it will get 31 bit storage. 3) Cleaned up the assembler support routine so it uses CMSCALL to issue CMS commands. The routine is still called as _SVC202_ from C defined as follows:- _SVC202_(PARMS,CODE,ERROR) WHERE :- PARMS IS A POINTER TO AN SVC202 PARAMETER LIST CODE IS A CODE TO SAY OF &CONTROL IS ON OR OFF ERROR is the RC from the CMSCALL. In addition there is some support for floating point. a. The s/f/printf, now has soem basic support for doubles(). b. strtod() works for numbers with a fractional part. c. There is a preliminary "alpha" release of support for "math.c" Sorting out the runtime to use natice CMS I/O is high on my ToDo list, along with cleaning up the implementation of math.h. I think math.h is next as "Cody and Waite" arrived on an inter-library loan so I need to finish it before I have to send it back:-) Hope this is all OK Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
I have an interest in this because there is no budget for additional software for z/VM or Linux. I tried a version of your GCCCMS from the VM/370 group on my z/VM 5.1 system on a z890 IFL and it worked fine. A thought for future development would be to fix the runtime library to use CMS native processes and maybe some CP/CMS interface routines (DIAG 8 type things). Thanks for all your hard work. /Tom Kern --- Dave Wade <[EMAIL PROTECTED]> wrote: > Thanks to the kind loan of access to a zVM system > from Dave Jones @ V-Soft Software I now have a GCC323 > that runs on VM, and generates "traditional" 370 > assembler. There is also a minimal run time library so > that CMS "modules" can be produced. Due to some of the > MVS macros used it still uses the code runs RMODE > 24/AMODE 31, but I am looking to fix this in the > medium term. > > Is this likely to be of interest to any one, as I > guess that most of you have IBMs C Compiler. If it is > what would be the best way to package it. I assume a > VMARC file would fine with most folks. > > Any other thoughts on this? > > Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
Rich, Should be a "Beta Release" available within the next few days. I'll put a note on here. Main slow down at present is pile of Timber the good lady wants converting to decking :-(. Dave. --- Rich Smrcina <[EMAIL PROTECTED]> wrote: > Sounds good! Please let us know where and when you > set it up for download. > > Dave Wade wrote: > > --- Rich Smrcina <[EMAIL PROTECTED]> wrote: > > > >> VMARC format works. What is minimal at this > point? > >> > > > > I guess thats a "moot point". It actually now has > > almost all of the ANSI standard routines in place > and > > the compiler will (almost) re-compile itself on > VM. A > > couple of routines fail, I think because they run > out > > of space below the 16Meg line and I am > investigating > > why... > > > > However to my mind the ANSI "C" library is some > what > > limited. So no SOCKETS, SHELL, DUP etc... > > > > The implmenetation also has some areas of > weakness. > > > > 1. Floating point support. > > == > > > > It still uses "tradional" floting point not IEEE.. > > > > Most of math.h is implemented but some of it is of > > limited accuracy and (almost) totally lacking in > error > > checking. > > > > s/f/printf() will output doubles, and strtod() > works > > (but again no error checking). scanf() has no > support > > for doubles(). I guess that if it does not use > > floating point and is single threaded it has a > good > > chance been made to work with minimal changes. > > > > I have actually built the BREXX interpreter from > > > > http://bnv.home.cern.ch/bnv/software/ > > > > on VM/370r6 and much to my suprise it seems to > work > > very well. I did need to implement some extra > routines > > (setjmp/longjmp) for this to work, but of course > they > > are now there for every one to use > > > > 2. File I/O. > > > > > > The run time uses OS file I/O, which I am not sure > I > > like. Despite breaking the rules by issuing > FILEDEF > > commands it seems to work fine. Again as its ANSI > > things like fcntl is not provided... > > > > 3. External routine names > > == > > > > I have been generating traditional TEXT decks and > > loading with "LOAD" and "INCLUDE" so limited to > > 8-chars on external routine names. > > > > Probably other stuff breaks as well. Its not had a > > real thrashing. On the plus side:- > > > > 1. All the source code is available, even the > (few) > > bits I wrote so you can always have a good giggle > at > > my "C" and Assembler :-) (or if feeling generous > > improve them) > > > > 2. The run time is pure public domain (compiler is > of > > course GPL) so you can do what you want with any > > MODULES you produce... > > > > 3. Its totally free. > > > > Dave. > > > > > > > >> Dave Wade wrote: > >>> Folks, > >>> > >>> Thanks to the kind loan of access to a zVM > >> system > >>> from Dave Jones @ V-Soft Software I now have a > >> GCC323 > >>> that runs on VM, and generates "traditional" 370 > >>> assembler. There is also a minimal run time > >> library so > >>> that CMS "modules" can be produced. Due to some > of > >> the > >>> MVS macros used it still uses the code runs > RMODE > >>> 24/AMODE 31, but I am looking to fix this in the > >>> medium term. > >>> > >>> Is this likely to be of interest to any one, > as > >> I > >>> guess that most of you have IBMs C Compiler. If > it > >> is > >>> what would be the best way to package it. I > assume > >> a > >>> VMARC file would fine with most folks. > >>> > >>> Any other thoughts on this? > >>> > >>> Dave. > >>> > >>> > __ > >>> Do You Yahoo!? > >>> Tired of spam? Yahoo! Mail has the best spam > >> protection around > >>> http://mail.yahoo.com > >>> > >> -- > >> Rich Smrcina > >> VM Assist, Inc. > >> Main: (262)392-2026 > >> Cell: (414)491-6001 > >> Ans Service: (360)715-2467 > >> rich.smrcina at vmassist.com > >> > >> Catch the WAVV! http://www.wavv.org > >> WAVV 2007 - Green Bay, WI - May 18-22, 2007 > >> > > > > > > __ > > Do You Yahoo!? > > Tired of spam? Yahoo! Mail has the best spam > protection around > > http://mail.yahoo.com > > > > -- > Rich Smrcina > VM Assist, Inc. > Main: (262)392-2026 > Cell: (414)491-6001 > Ans Service: (360)715-2467 > rich.smrcina at vmassist.com > > Catch the WAVV! http://www.wavv.org > WAVV 2007 - Green Bay, WI - May 18-22, 2007 > __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM and run time environment.
--- Rick Troth <[EMAIL PROTECTED]> wrote: > I wonder: could you get it to call LE libraries > instead of OS libs? I don't see why not, but I am not sure how much work would be required. The present code is pretty tied into OS DCBs and QSAM I/O. I had not considered using LE as I know very little about it, and its not available on the "old" VM/370 system that I started on. I was intending to change it to call the FSIO stuff as those routines are available on both old and new platforms, and I understand how they work. However this would reduce functionality as at present the code will read and write any device that you can FILEDEF (Tape/Spool/Console/CMS Dasd/OS DASD/SFS) and FSIO can only read CMS & SFS Dasd. (so I would have to add some code to read and write the console). However I will take a look at LE and see if it could be used. > > -- R; > Dave Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
Sounds good! Please let us know where and when you set it up for download. Dave Wade wrote: --- Rich Smrcina <[EMAIL PROTECTED]> wrote: VMARC format works. What is minimal at this point? I guess thats a "moot point". It actually now has almost all of the ANSI standard routines in place and the compiler will (almost) re-compile itself on VM. A couple of routines fail, I think because they run out of space below the 16Meg line and I am investigating why... However to my mind the ANSI "C" library is some what limited. So no SOCKETS, SHELL, DUP etc... The implmenetation also has some areas of weakness. 1. Floating point support. == It still uses "tradional" floting point not IEEE.. Most of math.h is implemented but some of it is of limited accuracy and (almost) totally lacking in error checking. s/f/printf() will output doubles, and strtod() works (but again no error checking). scanf() has no support for doubles(). I guess that if it does not use floating point and is single threaded it has a good chance been made to work with minimal changes. I have actually built the BREXX interpreter from http://bnv.home.cern.ch/bnv/software/ on VM/370r6 and much to my suprise it seems to work very well. I did need to implement some extra routines (setjmp/longjmp) for this to work, but of course they are now there for every one to use 2. File I/O. The run time uses OS file I/O, which I am not sure I like. Despite breaking the rules by issuing FILEDEF commands it seems to work fine. Again as its ANSI things like fcntl is not provided... 3. External routine names == I have been generating traditional TEXT decks and loading with "LOAD" and "INCLUDE" so limited to 8-chars on external routine names. Probably other stuff breaks as well. Its not had a real thrashing. On the plus side:- 1. All the source code is available, even the (few) bits I wrote so you can always have a good giggle at my "C" and Assembler :-) (or if feeling generous improve them) 2. The run time is pure public domain (compiler is of course GPL) so you can do what you want with any MODULES you produce... 3. Its totally free. Dave. Dave Wade wrote: Folks, Thanks to the kind loan of access to a zVM system from Dave Jones @ V-Soft Software I now have a GCC323 that runs on VM, and generates "traditional" 370 assembler. There is also a minimal run time library so that CMS "modules" can be produced. Due to some of the MVS macros used it still uses the code runs RMODE 24/AMODE 31, but I am looking to fix this in the medium term. Is this likely to be of interest to any one, as I guess that most of you have IBMs C Compiler. If it is what would be the best way to package it. I assume a VMARC file would fine with most folks. Any other thoughts on this? Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com -- Rich Smrcina VM Assist, Inc. Main: (262)392-2026 Cell: (414)491-6001 Ans Service: (360)715-2467 rich.smrcina at vmassist.com Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007 __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com -- Rich Smrcina VM Assist, Inc. Main: (262)392-2026 Cell: (414)491-6001 Ans Service: (360)715-2467 rich.smrcina at vmassist.com Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
Re: GCCCMS for zVM
I wonder: could you get it to call LE libraries instead of OS libs? -- R;
Re: GCCCMS for zVM
--- Rich Smrcina <[EMAIL PROTECTED]> wrote: > VMARC format works. What is minimal at this point? > I guess thats a "moot point". It actually now has almost all of the ANSI standard routines in place and the compiler will (almost) re-compile itself on VM. A couple of routines fail, I think because they run out of space below the 16Meg line and I am investigating why... However to my mind the ANSI "C" library is some what limited. So no SOCKETS, SHELL, DUP etc... The implmenetation also has some areas of weakness. 1. Floating point support. == It still uses "tradional" floting point not IEEE.. Most of math.h is implemented but some of it is of limited accuracy and (almost) totally lacking in error checking. s/f/printf() will output doubles, and strtod() works (but again no error checking). scanf() has no support for doubles(). I guess that if it does not use floating point and is single threaded it has a good chance been made to work with minimal changes. I have actually built the BREXX interpreter from http://bnv.home.cern.ch/bnv/software/ on VM/370r6 and much to my suprise it seems to work very well. I did need to implement some extra routines (setjmp/longjmp) for this to work, but of course they are now there for every one to use 2. File I/O. The run time uses OS file I/O, which I am not sure I like. Despite breaking the rules by issuing FILEDEF commands it seems to work fine. Again as its ANSI things like fcntl is not provided... 3. External routine names == I have been generating traditional TEXT decks and loading with "LOAD" and "INCLUDE" so limited to 8-chars on external routine names. Probably other stuff breaks as well. Its not had a real thrashing. On the plus side:- 1. All the source code is available, even the (few) bits I wrote so you can always have a good giggle at my "C" and Assembler :-) (or if feeling generous improve them) 2. The run time is pure public domain (compiler is of course GPL) so you can do what you want with any MODULES you produce... 3. Its totally free. Dave. > Dave Wade wrote: > > Folks, > > > > Thanks to the kind loan of access to a zVM > system > > from Dave Jones @ V-Soft Software I now have a > GCC323 > > that runs on VM, and generates "traditional" 370 > > assembler. There is also a minimal run time > library so > > that CMS "modules" can be produced. Due to some of > the > > MVS macros used it still uses the code runs RMODE > > 24/AMODE 31, but I am looking to fix this in the > > medium term. > > > > Is this likely to be of interest to any one, as > I > > guess that most of you have IBMs C Compiler. If it > is > > what would be the best way to package it. I assume > a > > VMARC file would fine with most folks. > > > > Any other thoughts on this? > > > > Dave. > > > > __ > > Do You Yahoo!? > > Tired of spam? Yahoo! Mail has the best spam > protection around > > http://mail.yahoo.com > > > > -- > Rich Smrcina > VM Assist, Inc. > Main: (262)392-2026 > Cell: (414)491-6001 > Ans Service: (360)715-2467 > rich.smrcina at vmassist.com > > Catch the WAVV! http://www.wavv.org > WAVV 2007 - Green Bay, WI - May 18-22, 2007 > __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Re: GCCCMS for zVM
VMARC format works. What is minimal at this point? Dave Wade wrote: Folks, Thanks to the kind loan of access to a zVM system from Dave Jones @ V-Soft Software I now have a GCC323 that runs on VM, and generates "traditional" 370 assembler. There is also a minimal run time library so that CMS "modules" can be produced. Due to some of the MVS macros used it still uses the code runs RMODE 24/AMODE 31, but I am looking to fix this in the medium term. Is this likely to be of interest to any one, as I guess that most of you have IBMs C Compiler. If it is what would be the best way to package it. I assume a VMARC file would fine with most folks. Any other thoughts on this? Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com -- Rich Smrcina VM Assist, Inc. Main: (262)392-2026 Cell: (414)491-6001 Ans Service: (360)715-2467 rich.smrcina at vmassist.com Catch the WAVV! http://www.wavv.org WAVV 2007 - Green Bay, WI - May 18-22, 2007
GCCCMS for zVM
Folks, Thanks to the kind loan of access to a zVM system from Dave Jones @ V-Soft Software I now have a GCC323 that runs on VM, and generates "traditional" 370 assembler. There is also a minimal run time library so that CMS "modules" can be produced. Due to some of the MVS macros used it still uses the code runs RMODE 24/AMODE 31, but I am looking to fix this in the medium term. Is this likely to be of interest to any one, as I guess that most of you have IBMs C Compiler. If it is what would be the best way to package it. I assume a VMARC file would fine with most folks. Any other thoughts on this? Dave. __ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
