toby@gargoyle.UUCP (Toby Harness) (02/07/86)
styborsk@hpspkla.UUCP (styborsk) writes: > >Regarding the various kinds of "hp9000", you are right, there is more than >one type... Specifically, there are two classes of hp9000's that I am aware >of currently. The first class consists of hp9000 series 300 (the 310 and the >320) and the older hp9000 series 200 (237, etc). There is also the hp9000 >series 500. In there first class, any computer running HP-UX 5.x should look >fairly identical to any program that doesn't specifically look for trouble. >In fact, both classes should be fairly identical. The major differences (from >a software environment point of view) that I am aware of are: 1) on the s500 >you cannot dereference a null pointer successfully (which I believe is correct >anyway) and 2) on the s500, the stack grows in a different direction... It would not be unfair to say GNU emacs is looking for trouble. The series 500 has a highly segmented architecture - 1 stack seg, up to 1023 code segs, 1 direct data seg, and 1 (paged or unpaged) indirect data seg per process. Considering the task that 'exec' has to load an a.out file in to all of this, it would be distinctly non-trivial to write an 'unexec' - GNU emacs is pretty hard to live with without resorting to this type of stunt. And as of several versions ago, GNU emacs still made numerous improper assumptions about the way memory was handled. I'm sure (I hope) most of these have have been corrected. Unexec is more than non-trivial if you can't get the HP-UX/SUN internals and FOCUS (the chipset) docs out of HP, which, last I heard, was not very likely. Hey HP - you talking yet? Toby Harness Ogburn/Stouffer Center, University of Chicago ...ihnp4!gargoyle!toby