[Lldb-commits] [PATCH] Support for running test suite using makefiles on Linux

[Peter Collingbourne]

On Sun, Jun 19, 2011 at 01:56:06PM -0600, Charles Davis wrote:
> > [From patch 1]
> > > @@ -31,15 +36,27 @@ endif
> > [...]
> > > +ifeq "$(DYLIB_NAME)" ""
> > You do mean "ifneq" here, right? Otherwise, you'll get back lib.so (or
> > lib.dylib) as the shared library's filename. :)

Yes, I do :).  Fixed.

> On 6/18/11 10:58 PM, Charles Davis wrote:
> > On 6/18/11 6:32 PM, Peter Collingbourne wrote:
> > [From patch 4]
> >> @@ -35,21 +35,23 @@ LLDB_SWIG_INCLUDE_DIRS += -I"/usr/include"
> >>  LLDBWrapPython.cpp:
> > [...]
> >> +	$(Verb) $(RM) -f $(PYTHON_DIR)/_lldb.so
> >> +	$(Verb) $(AliasTool) $(LIBLLDB) $(PYTHON_DIR)/_lldb.so
> > This can't work. At least, it can't if the $(AliasTool) expects the
> > source file to exist. At first, I thought you'd need to depend on
> > liblldb.so; but that won't work either, because liblldb.so depends on
> > LLDBWrapPython.cpp! You'll have to make the _lldb.so symlink elsewhere.
> Never mind. ln -s works even if the source file doesn't exist, even on
> Darwin. But it wouldn't work on Windows, where the $(AliasTool) is not
> 'ln -s' but 'cp'/'copy'. But that's OK for now, because LLDB doesn't
> even work on Windows :).

I can't see any assignment to $(AliasTool) in the makefile system other
than "ln -s".  So I assume that on Windows these commands would end
up using whatever the MinGW/MSYS/Cygwin emulation of symlinks are,
which (if they are POSIX compliant) should deal with nonexistent
source files.

Thanks,
-- 
Peter