2009/1/1 John Malmberg <wb8...@gmail.com>: > John Malmberg wrote: >> Here is a first pass at a make_patchnum.com. > > Ignore that file, the arguments need extra quoting. > > New file will be coming next year.
Hi guys. Im sorry to have done this to you, but Ive broken VMS bleadperl again. :-( That is, I've (with help from others) converted the make_patchnum.sh to a perl script, and reworked the *nix build process so it now uses miniperl to build. When it is building miniperl the first time it uses a precanned header file. The idea of this is that Win32 and VMS and *NIX will be able to share the same script. It looked like Win32 wasnt going to be so easy to handle otherwise, and i felt that the reduction in maintenance justified it. You can see the overall diff here: http://perl5.git.perl.org/perl.git/commitdiff/505afc734ed?hp=7b1668ed The key points: * .patchnum is eliminated and special defines are no longer passed in by cflags when compiling perl.c, instead all the defines are now located in a single header file (formerly unpushed.h, now git_version.h). No more calls to awk. * unpushed.h renamed to git_version.h, and a new file "stock_git_version.h" which is copied to git_version.h when making miniperl the first time. Once miniperl is available make_patchnum.pl is executed, updating git_version.h (and lib/Config_git.pl) * the makefile aspect of things has changed a bit, now if the git_version.h header is updated a recursive make is started of miniperl. This means that perl.c and miniperlmain.c and miniperl itself are recompiled/rebuilt an additional time, but also means that the interesting logic can now be implemented in perl code. Real sorry about breaking things, I wish I was in a position to help you guys fix up the VMS build process. However i think once you adapt to this change there should be no more need for VMS specific changes, everything can be handled in perl code. Cheers, yves -- perl -Mre=debug -e "/just|another|perl|hacker/"
