Hrvoje> Raymond Toy <toy(a)rtp.ericsson.se&gt; writes: Hrvoje> Maybe, maybe not. Remember that it could also be a memory corruption Hrvoje> bug, and those can go away when you change *anything* -- and come back Hrvoje> to haunt you a year later. Yes. The reason I think it's a gcc problem is that start1 is set to BUF_BEGV(curent_buffer) = 1 in Fwrite_region_internal. However, at some later point it magically gets changed to 8388xxxx in the call to make_lisp_buffer_stream. A peek at the code shows that start1 is only set in a few places and none of them can do this. Recompiling with Sun CC seems to have taken care of this problem. But I can't conclusively say it's not a memory corruption problem. Hrvoje> Is gcc 2.95.2 a stable release of Gcc? Well, it's definitely not a snapshot, and it's on the cygnus gcc web site as being a release. Ray