[Info-vax] Looking into C-include files on VMS

[John E. Malmberg]

Joerg Schilling wrote:
> In article <alpine.LFD.2.00.0910281551350.24799 at libra.gmcl.internal>,
> Rob Brown  <mylastname at gmcl.com> wrote:
> 
>>> I would need aprox. 60 MB for unpacking and compiling.
>> Type SHOW QUOTA at the prompt.  If you don't have enough, then send 
>> MAIL to ASSIST and explain why you need more.  Usually they grant it 
>> for projects like yours.
>>
>> Or maybe the "scratch space" facility will be adequate for your needs.
> 
> I did create an account but it seems that there is no GNV installed, as 
> I am unable to call bash and similar programs. There is "vim", there is
> zip but no tar....
> 
> I'll see...

eisner.decus.org is the same as eisner.encompasserve.org and the can be 
reached as encompasserve.org.  The maintainers of eisner own the 
encompasserve.org domain, the decus.org domain is owned by the 
organization formerly know as Decus.

If you will look at my original post, it shows where the ODS5 volume is: 
usr_ods5:

$ @gnv$gnu:[lib]gnv_setup.com

Note that it is gnv$gnu, not gnu.  The gnu name violates VMS logical 
name standards for an installed product.  (That fix has been submitted 
to HP, but I do not know when it will be incorporated)

$ bash
bash$ cd /usr_ods5/malmberg/
bash$ ls
SCHWERN-TEST-MORE.TGZ      TEST-MORE      apr-iconv-1.2.1
SCHWERN-TEST-MORE.TGZ_OLD  TEST-MORE_OLD
bash$ tar --version
tar (GNU tar) 1.19
Copyright (C) 2007 Free Software Foundation, Inc.
Written by John Gilmore and Jay Fenlason.



bash$ vim --version
VIM - Vi IMproved 6.3 (2004 June 7, compiled Dec 10 2007 16:56:30)
OpenVMS version
Included patches: 1-28

One of the main features of eisner.encompasserve.org is a moderated 
notes conferences that is invoked by the command NOTES.

The "PORTING_TO_VMS" conference is not there by default.  It is added 
with the command "ADD ENTRY PORTING_TO_VMS", or you can use web access 
that I posted the URL for earlier.

When posting in the notes conferences, please make sure that the lines 
are only 80 characters long.  The terminal interface to notes does not 
automatically wrap the lines.

Joerg Schilling wrote:
 > In article 
<a26ade9d-4fb1-4ecb-9b84-42460674c4f8 at s15g2000yqs.googlegroups.com>,
 > Steven Schweda  <sms.antinode at gmail.com> wrote:
 >> Joerg Schilling wrote:
 >>
 > On some platforms, GNU make will work also but the GNU make Makefile
 > parser has some bugs that tends to cause problems on platforms that
 > use CR/LF in files and backslash in filenames.

The VMS native text files are counted records, VMS also supports CR/LF 
and LF endings.  Access through the VMS CRTL in binary gives you the 
native format, and opening as a text file, VMS will make the file look 
like it has LF line endings.

 >>   Trying to force-fit a UNIX-like "configure" script onto a
 >> VMS system seems to me to be the wrong way to do this job.
 >> Which is why I did it differently.  UnZip and VMSTAR both
 >> adjust file names as needed for an ODS2 file system.  My VMS
 >> builders seem to cope, so far as I know.

As I understand it the translation is to replace non-ods2 characters 
with underscores.

 > Why? This is something that was created to test the features.

Configure makes incorrect assumptions about the build environment on any 
platform that maintains backwards binary compatibility with older releases.

A C program built on VMS 5.5-2 (Earliest current C compiler supports) 
will operate the same way on VMS 7.3, the last VAX release.

See my other reply to you.

 > What I
 > did see on the VMS account I aquired a week ago looks like something
 > that is very close to a working solution and HP did even claim that
 > on a newer system configure should work now.

I am the one that put the changes into GNV bash that eliminates most of 
the problems with running configure.  That requires VMS 8.2 at a minimum 
and the latest GNV kit, or one that has my patches applied.

I am very much aware of what limitations remain.

 >>> -	OpenVMS 7.0 should support a POSIX compliant struct stat.st_ino
 >>> 	but I cannot make this work.

Nope, you need VMS 8.2 as I posted earlier.

 > The current configure I am using has aprox. 650 individual tests.
 > If you miss results for tests that would succeed, the software will
 > not use features that could be available.

It usually takes me several passes to get the configure script to 
produce the correct output for VMS.

I had to put back doors in the GNV CC wrapper program that allow me to 
trick the configure script into coming up with the correct answer for VMS.

Most of these hacks could be eliminated if configure was made to test 
the functions with the header files in place instead of testing 
functions with out header files.  It would also shorten the runtime of 
the configure program, since many of the configure scripts eventually 
run that tests.  See my previous reply.

GNV is not a perfect UNIX emulation on VMS, it is a work in progress.

I really recommend that you look over the porting_to_vms conference that 
I posted earlier.  While you may be able to get configure scripts to run 
with out doing so, it will be harder.

-John
wb8tyw at qsl.network
Personal Opinion Only