Date: Thu, 11 Apr 1996 14:18:00 +0200 (IST) From: Eli Zaretskii To: Morten Welinder Cc: djgpp-workers AT delorie DOT com Subject: Re: Emacs works under v2.0! In-Reply-To: <199604110942.LAA14115@tyr.diku.dk> Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Thu, 11 Apr 1996, Morten Welinder wrote: > Aha! That was why my alloca-based hack happened to solve the > problem. Yes. But even after I understood why, I was reluctant to define large automatic variables, because as you remember, the stack that the pure temacs uses just wastes the disk space in the dumped binary. Right now, src/Makefile stubedits temacs to 40K stack (and that includes about 15K spare space, just in case, I think 25K is enough). I think we should discuss the possible variants of the improved `write' for the next release. For example, my quick-and-dirty work-around in src/msdos.c defines a 64KB static buffer, but I don't think this is appropriate in the library version. > It ticked into the source yesterday. Good work! No, that were the old changes that let you build Emacs without `ralloc'. The last portion of the changes that solves the `ralloc' problem was sent only this morning. Btw, when you have time to build and play with it, please look at the speed it consumes characters. It seems to me that if you continuously press a key that moves the cursor, it works slower than the v1.x version. It's best seen when lazy-lock and paren are both loaded: sometimes Emacs won't scroll the display when the cursor hits the edge of visible part of the buffer. One thing that I thought of is that Emacs uses `int86': aren't they supposed to be slower than `__dpmi_int'?