From: Martin D. <mdo...@mu...> - 2001-11-16 20:38:38
|
Laurent Vogel wrote: > The problem is, that when adding both language tables, the > TEXT segment goes so big, that it ends AFTER the START of > the DATA segment. This can be detected by doing "make dsm" > and then "grep __etext map". > > Apparently in this case the start of DATA clobbers the > end of TEXT! For me the fact that ld does not report this > situation is a bug of the cross-mint tools. But what can we do ? > > I don't see any practical way of asking gcc to simply put data > at the end of the TEXT section. We simply have to remember to > check the __etext address whenever there is a bug. I can even > create a Makefile entry to just display the memory map if it > can help. Urgs! Bad thing. Will not gcc put the data segment after the text one automatically? Well, you said not. That's bad. Frank - any experiences (you are always good for experiences ;-) Ok, we could simply put the data segment to a higher address. Or we make EmuTOS 256 kb big, like the 2.0x TOSs. Good thing. that you at least found this bug - such things are hard to find. I once did fight against a too small stack... -- Martin |