------- Comment #4 from pinskia at gcc dot gnu dot org 2005-10-28 00:24 ------- (In reply to comment #3) > I wouldn't call this a regression; I don't think there's any guarantee that > unused statics stay around, even with -O0.
It is a regression as turning on unit-at-a-time at -O0 changed it. > However, I tend to agree that it would be better if they did stay around at > -O0. In other words, I wouldn't be opposed to adding such a guarantee. > Do we know if this is a front-end problem or a cgraph problem? I think this is a cgraph issue but that is only because nobody thought unit at a time would be enabled at -O0 when it was first made. > Is cgraph > throwing stuff away at -O0, or are we not telling cgraph about the variable? cgraph is throwing things away when unit-at-a-time is enabled which is all the time with C++ in 4.0 and above. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24561
