Tabelle 9.1: Currently used platforms with VMake
VMake will be soon adapted to the IRIX operating system from SGI to allow compilation of all our programs on this platform. Similar to the ``osf'' configuration this will be a 64bit configuration. Since all the software is already checked for 64bit cleanness of integer and pointer mismatches, this should be quite simple as was shown for other UNIX derivates.
Within VISTA quite a lot of code is automatically generated by either UNFUG or TAC. YACC and LEX are replaced within VISTA by their free implementation from GNU. Table 9.2 shows projects using the different automatic code generators of VMake and the amount of generated code. For UNFUG and TAC the numbers are measured in lines of codes. In addition for TAC the number of functions and constants generated using the language binding mechanism is shown.
Tabelle 9.2: Automatically generated code in VISTA
For the PifGridSupport project, the code is generated using the Voops-Target for UNFUG which is a subclass of the Unfug-Target. UNFUG is currently also used to solve the problem of instancing template C++ instances. The automatic language binding for VLisp is generated usually in the project of the new VLisp interpreter whereas the FORTRAN bindings are generated as objects for inclusion in a library.
Currently VMake is used to build and test these projects under two VMS and seven UNIX derivates. The selftest is done automatically every night to check whether the CVS repository can be built and the programs are functional. This allows to update a global installation for multiple platforms and configurations on an almost daily base. To have a good backup, the last released version of each software is stored globally. Using most of the software from the global installation cuts down local disc usage per developer significantly. A second advantage is that one can concentrate on the own work.