Re: C vs C++ compilations

[Gary Thomas <gary at mlbassoc dot com>]

On Wed, 2003-10-15 at 13:43, Jonathan Larmour wrote:
> John Dallaway wrote:
> > 
> > 
> > Long-term, it seems that we will definitely need to separate C and C++ 
> > build flags. We could certainly introduce CYGBLD_GLOBAL_CXXFLAGS 
> > (similar to CYGBLD_GLOBAL_CFLAGS) and modify the makefile generation 
> > code to use it. This would not be a huge task, but modifying 100+ 
> > platforms HALs to accommodate the change might be. I'm not so keen on 
> > the option of hacking the makefile generation code to treat 'legacy' 
> > platform HALs differently.
> 
> See bug 1000035. The long term solution definitely doesn't involve 
> anything like a CYGBLD_GLOBAL_CFLAGS. Flags to deal with warnings, 
> optimisation, debugging etc. should be brought out separately - there's no 
> reason for most of them to be anywhere near a platform HAL.
> 
> >> The attached version [ of the patch ] works fine on my
> >> PowerPC and ARM targets.  There still are some warnings when there are
> >> special rules.
> >>
> >> What version (date) of ecosconfig has the support for the full flag set?
> >>
> >> I realize that it will be some effort to split up the CDL and handle
> >> these problems fully, but as I said I think it's something we should
> >> address.  The patch could be committed now as it doesn't conflict with 
> >> any other changes (e.g. using CXXFLAGS)
> > 
> > 
> > It seems that your proposed patch constitutes a temporary workaround to 
> > silence build warnings before we address the real problem. A couple of 
> > comments:
> > 
> > a) Why do you modify the flags for building both C and C++ files in your 
> > patch? From your description of the problem, it seems that the flags you 
> > cite need only be removed when compiling C files.
> 
> More interesting is why the patch worked even despite that: -fvtable-gc 
> should be removed globally anyway - last I knew the GCC support was broken 
> (and when it worked, it was apparently unreliable anyway, and in any case 
> no core eCos code uses it). -Woverloaded-virtual is just a warning. I'm 
> surprised compiles didn't fail with -fno-rtti missing though. Maybe 
> luckily enough the right sections are present and correctly defined in the 
> chosen targets/architectures. But yes the patch should be changed if it 
> affects c++ compiles too.
> 

The patch is doing the right thing - it only affects .c (and currently 
.S) files, not .cxx files.  I know the patch looks like it does mess 
with .cxx files, but I've checked the result.

> > However, I'm wondering whether we should just bite the bullet and split 
> > out the C and C++ build flags soon rather than introduce a workaround. I 
> > would be interested in comments from the other maintainers.
> 
> Again bug 1000035. It's tied up with the makefile generator rewrite.
> 
> Jifl
-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates