[tree-ssa] Mainline merge plan

[law@redhat.com]

In message <200402251550.i1PFon120368@pc960.cambridge.arm.com>, Richard Earnsha
w writes:
 >> >>>>> Richard Earnshaw writes:
 >> 
 >> Richard> I've no idea what mudflap is supposed to be for, so it's unclear t
 >o me 
 >> Richard> whether it's an important part of the compiler or just another 
 >> Richard> nice-to-have bolt on.  The following patch just disables it from t
 >he build 
 >> Richard> and does seem to solve the immediate problem of building the libra
 >ries.  
 >> Richard> Whether or not this is the correct approach somebody else will hav
 >e to 
 >> Richard> judge.  A better patch would probably move the gcc thread-detectio
 >n code 
 >> Richard> to the top level and use that do decide what the target capabiliti
 >es were.
 >> 
 >> 	Mudflap: narrow-pointer bounds-checking by tree rewriting.
 >> 
 >> 	target-libmudflap used to build on AIX, but fail in the testsuite.
 >> It recently began failing to build.  It is not a critical component of
 >> Tree-SSA.
 >> 
 >> David
 >
 >If it's not critical anywhere, then I'm inclined to suggest that it should 
 >be temporarily disabled everywhere for now, in the interests of sorting 
 >out the main issue -- the merge of tree-ssa itself.  Once that is sorted 
 >out we can go back to looking at mudflap.
Or have it disabled by default and allow targets where it is supported
to explicitly enable it.
jeff