Date: Mon, 21 Oct 1996 18:42:17 +0200 (IST) From: Eli Zaretskii To: Howard Kaikow Cc: djgpp AT delorie DOT com Subject: Re: HK: v 2.1 In-Reply-To: <1.5.4.32.19961020202851.0066bbe8@mv.mv.com> Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Sun, 20 Oct 1996, Howard Kaikow wrote: > At 06:41 PM 10/20/96 +0200, Eli Zaretskii wrote: > > > >Are these .exe files stripped? If not, please compare the sizes of the > >stripped binaries (the size differences might be caused by debugging > >symbols). > > I have previously sent the command line that was used to compile. And the > same command line was used with v2 as with v2.1 so we ar ecomparing oranges > with oranges. Yes, but if most of the difference is due to more debugging symbols being output by GCC 2.7.2.1, does this really matter? The symbols don't belong to the executable code (and aren't loaded at runtime), and most DJGPP programs are distributed without the symbols. v2.01 executables indeed are somewhat larger than those produced by v2.0, but by much less than 30KB. I've just built a couple of test programs to see the difference, and here are the results: v2.0 v2.0 stripped v2.01 v2.01 stripped hello.c 57K 32K 79K 36K systest.c 90K 57K 143K 68K `hello.c' is the classic ``Hello, world!'' program, where most of the additional 3K is in the startup code (better support for LFN, filename globbing on Windows 95, etc.) and in `printf'. `systest.c' is a program which calls `system', where *a lot* of functionality was added (read the library docs, there are too many changes to list here). Conclusion: stripped binaries in v2.01 should be between 3KB and 12KB larger than in v2.0; but the unstripped binaries are *much* larger, probably due to more debugging symbols being emitted by the compiler/assembler.