On Mon, Jul 25, 2005 at 04:48:45PM -0400, DJ Delorie wrote: >> Presumably, people with blanket write privs and people responsible for >> the build machinery. > >Yup, that's them. > >I did a little historical digging on this item, and the original >trigger was http://gcc.gnu.org/ml/gcc-patches/2005-05/msg00280.html >where Paolo needed to switch from symlinks to hardlinks for toplevel >bootstrapping. So, I think we've confused the issue enough that we >may be making bad decisions (or at least *I* am confused and may make >bad decisions ;). > >Paolo, could you go back and think about the bootstrapping problem >from MinGW's perspective? > >Danny, could you summarize which build features MinGW supports? (I >assume cp and $(SHELL) but not symlinks, hardlinks, or #!). DJGPP and >Cygwin shouldn't be a problem here; they both support symlinks and #! >but I don't think they support hardlinks.
For shell commands, the MSys build environment supports pretty much everything that Cygwin does, which would be everything currently in that makefile rule. The problem is that when gcc executes, it uses (eventually) the unadulterated Windows CreateProcess call which does not understand cygwin/MSys symlinks or '#!' shell scripts or any other type of shell script. It would understand a .bat or .cmd file, but adding .bat or .cmd file processing to Makefile.in is probably not something that we want to do. Hard links would be supported by everything since Windows understands hard links at the OS level and Cygwin/MSys can create them (in some cases). They aren't a generic solution, however, as the makefile rule demonstrates. Maybe one solution would be to patch pex-win32 for mingw so that it could understand '#!' style shell scripts? That would at least allow bootstrapping. It wouldn't be useful generic pure-windows installation, but presumably a user wouldn't have to worry about that because everything would be installed correctly. cgf
