From: antispam@fricas.org   
      
   Simon Clubley  wrote:   
   > On 2025-08-30, Waldek Hebisch  wrote:   
   >> Simon Clubley  wrote:   
   >>>   
   >>> That's where I gave up. I either didn't understand the required   
   >>> VMS-specific build steps or there were bits missing from the public   
   >>> source kits.   
   >>   
   >> There are missing pieces, build needs VMS header files and libraries.   
   >> If those pieces are incomplete, then configuration may go wrong   
   >> and lead to trobles later.   
   >>   
   >> Currently I have fairly incomplete set of header files, so I expect   
   >> trouble due to this.  OTOH libgcc seem to compile fine, so C   
   >> compiler probably will work.  I worked around trouble with   
   >> shared images by configuring gcc with '--disable-shared',   
   >> suboptimal but must do for now.   
   >>   
   >   
   > If you look at the links Arne posted, you will see I took across to   
   > Linux the VMS-supplied header files inc addition to the libraries.   
   > According to those notes, it appears I also had to write patches   
   > for those headers.   
      
   I saw Arne message only after writing message above.  My approach   
   differs, namely I use headers that I created.  They have minimal   
   content, basically what is needed to compile libgcc and a few other   
   things.   ATM I am working at creating needed libraries.  My   
   understanding is that for linking purposes I do not need any real   
   code in the libraries, just function names (in correct order).   
      
   I hoped that linking could be done without any libraries (this is   
   possible on most platforms), but VMS ld in binutils insists on having   
   libraries and up to now I have no work around for this.   
      
   > When I talked about bits missing, I wasn't thinking about the headers   
   > and libraries. Given that Adacore didn't start using Newlib or something   
   > similar to support VMS, that was pretty much expected.   
   >   
   > I was thinking more about bits missing within the compiler or binutils   
   > itself, especially given the crashes I had with Fortran. I was wondering   
   > if, for example, there are bits missing in the image initialisation   
   > sequence, along with other similar types of missing bits.   
      
   Do you have any indications that gcc Fortran or C++ worked on VMS say 25   
   years ago?  Presumably Ada Core cared mostly about Ada.  To build Ada   
   compiler they needed to build C compiler first.  I do not know if they   
   ever supported on VMS gcc version recent enough to require C++ to build.   
   If C++ and/or Fortran was never supported, then missing pieces in runtime   
   support are expected.  Also, there may be new pieces of platform support   
   needed by newer compiler versions, that nobody bothered to add for VMS.   
      
   BTW: I compared VMS specific headers in binutils-2.21 and binutils-2.43.   
   They are the same, except for some comment changes.  OTOH headers seem   
   to match reasonably well with file structure.   
      
   --   
                                 Waldek Hebisch   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)