On Fri, May 21, 2010 at 11:19 AM, Jeremy Huddleston <[email protected]> wrote: > The analyzer is correct. It sees the call to miFillPolyHelper on line > 1849 and assumes that pGC can change because it is not const ... > > My guess is that applying "const" correctly in many places will help > the SA avoid false positives like this.
Ooh, interesting. OK. miFillPolyHelper can't take a const pGC though, because eventually it passes it to ChangeGC (although with the invariant that it will be changed back before returning). What about this case though? http://people.freedesktop.org/~jeremyhu/clang/2010-05-20-1/report-bKYbbq.html Is clang having some kind of alias analysis trouble there? pDrawable isn't being passed to any functions, but &((WindowPtr) pDrawable)->borderClip is... Say, looks like this putative null-pointer dereference should be fixable by making AtomError _X_NORETURN: http://people.freedesktop.org/~jeremyhu/clang/2010-05-20-1/report-0TXsTP.html I'll send that patch along shortly. Jamey _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
