Xref: news-dnh.mv.net comp.os.msdos.djgpp:3109 Path: news-dnh.mv.net!mv!news.sprintlink.net!news1.is.net!news1.i1.net!scratchy.mi.net!netline-fddi.jpl.nasa.gov!hudson.lm.com!news.math.psu.edu!chi-news.cic.net!newsfeed.internetmci.com!howland.reston.ans.net!Germany.EU.net!news.dfn.de!RRZ.Uni-Koeln.DE!wmwap1.math.uni-wuppertal.de!schoenep From: schoenep AT wrcs1 DOT urz DOT uni-wuppertal DOT DE (Van Schoenepauck) Newsgroups: comp.os.msdos.djgpp Subject: Problem: Make eats up XMS Date: 25 Oct 1995 13:17:33 GMT Organization: University of Wuppertal Lines: 42 Nntp-Posting-Host: wrcs3.urz.uni-wuppertal.de To: djgpp AT sun DOT soe DOT clarkson DOT edu Dj-Gateway: from newsgroup comp.os.msdos.djgpp Hello *.*, just when I thought I had successfully installed DJGPP (vers. 1.12m4) _without_ a lot of trouble, I ran into the first problem: Make eats my memory! I tried to compile a test file using a rather simple Makefile of the form: xyz.exe: xyz.cc gcc -O2 -g xyz.cc -o xyz -lgpp coff2exe xyz When I feed this to Make, go32 (given the meminfo option) tells me upon start that it has about 3 mb of free XMS. But as soon as make spawns another copy of go32 which executes gcc, the new copy reports to have only -4 (yes, *minus* four) kb of free XMS, which of course results in some heavy swapping during compilation. In fact, I waited 20 minutes or so before I aborted, and then tried the same compilation from the command line, without Make: It took less than 50 seconds, because gcc used the whole 3 mb XMS for the compilation. I *do* have enough swapspace, so it should be no problem for go32 to page out the whole XMS to disk. And I *did* read the FAQ, where I could only find hints on what to do if there is not enough conventional mem. Now my questions are: 1. Why does Make use up XMS? I understand that every copy of go32 which is currently loaded takes up some of the conventional memory, but shouldn't it free as much high memory as possible for every child it spawns? 2. Is there a way around this? I don't want to do without Make if it comes to managing more complex programming projects, but a slowdown of this magnitude is, of course, unacceptable. Is it a bug? A feature? If anybody has a solution, *please* e-mail, or post a followup. Thanks in advance, Jan Schoenepauck schoenep AT wrcs3 DOT urz DOT uni-wuppertal DOT de