Xref: news2.mv.net comp.os.msdos.djgpp:1273 From: wald AT theory DOT lcs DOT mit DOT edu (David Wald) Newsgroups: comp.os.msdos.djgpp Subject: Re: Problems with cc1.exe (djgpp v 2.0) Date: 19 Feb 1996 22:26:06 GMT Organization: Theory of Computation, LCS, MIT Lines: 24 Message-ID: References: <4fumok$pr7 AT news DOT cea DOT fr> <4funbp$1uo AT news DOT uni-c DOT dk> <3123d1b7 DOT sandmann AT clio DOT rice DOT edu> <4g9hdb$ene AT news DOT uni-c DOT dk> NNTP-Posting-Host: woodpecker.lcs.mit.edu In-reply-to: jesper@chaos.fys.dtu.dk's message of 19 Feb 1996 09:54:51 GMT To: djgpp AT delorie DOT com DJ-Gateway: from newsgroup comp.os.msdos.djgpp In article <4g9hdb$ene AT news DOT uni-c DOT dk> jesper AT chaos DOT fys DOT dtu DOT dk (Jesper Skovhus Thomsen) writes: >Charles Sandmann (sandmann AT clio DOT rice DOT edu) wrote: >> So, either GCC 2.6.3 used a lot less stack than 2.7.2 does, or the beta >> testers don't compile huge C++ files, or maybe a bit of both. > >I cant say about C++ because I "only" uses C. However on might argue that >it is bad programming style to have a single huge C file insted of a >series of smaller C files that are linked together (I was able to compile >both libjpeg and libtiff without any [compiler] problems). I'll buy that, but the problem with DJGPP 2.0 C++ (at least my problem with it) isn't really "huge C++ files". It's that with template types the type information required by headers can turn a short, reasonable source file into a *lot* of work (and apparently, a nasty amount of memory allocation) for the compiler. The files I was (possibly still am -- I'm getting mixed results) having trouble with were quite reasonable-sized pieces of a much larger project. -David -- ============================================================================ David Wald http://theory.lcs.mit.edu/~wald/ wald AT theory DOT lcs DOT mit DOT edu ============================================================================