* We must use the full path to the source file when calling MSVC so that debugging information shows up in the object files.
* Because of cygwin's use of /cygdrive when using full paths, we must use a cygwin-wrapper script in some cases to allow native win32 programs to use a fully qualified cygwin path.
* We have to call configure using the full path so that $(srcdir) contains the
full path in the Makefiles. The rules have been changed for win32 so that it
always use $(srcdir)/$*.{c,cpp} as the sourcefile name even when the file is in
the cwd. This works around both the /cygdrive issue and the msvc's path info
optimization.
* We still use the wrapper when calling rc.exe & nsinstall. nsinstall
frequently takes multiple args and is called all over the tree so changing each
calling site is going to be expensive. (I'll have to check the logs again but
I think the rc wrapping can go.)
* nspr & ldap still use the wrappers for all native win32 progs. Neither nspr
nor ldap uses the acoutput-fast.pl script to speed up the substitution of
@srcdir@ into their Makefiles so, makefile substitution will break if we use
dos-paths.
* In the handful of directories where we copy srcfiles from another directory
in lieu of using VPATHs, we have to now copy those files to $(srcdir) so that
the default rules can find them.
Bug #141834 r=leaf,wtc,dmose,mcs
r=cls, r=pinkerton, a=asa
fix up packaging scripts so they can run from any directory
move from perl's "use" which is compile-time to require/import which are run-time, so that we can use the FindBin package.
also fix a typo which allows mozLock.pm to recognize windows paths
r=vishy. verified this solution works for mozilla, commercial and shrimp.
dont have bug number handy but this should be world++. basically from
now on, jar.mn rules will both create the jar and publish the flat files
to dist. the flat files are needed so that installers can compose the
files into a new naming system or delete some or whatever.