Date: Wed, 1 Nov 2000 17:11:16 +0200 (IST) From: Eli Zaretskii X-Sender: eliz AT is To: Rudolf Polzer cc: djgpp AT delorie DOT com Subject: Re: Internal compiler error - not covered by the FAQ In-Reply-To: <8tp2a8$nfv$1@nnrp1.deja.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Reply-To: djgpp AT delorie DOT com Errors-To: nobody AT delorie DOT com X-Mailing-List: djgpp AT delorie DOT com X-Unsubscribes-To: listserv AT delorie DOT com Precedence: bulk On Wed, 1 Nov 2000, Rudolf Polzer wrote: > > This sounds awfully like a hardware problem; see section 6.6 of the > > FAQ for more details. > > I can compile a Linux kernel and there should be a h/w problem? Why not? The memory layout on Windows and on Linux is very different; it's highly probable that on Windows GCC runs in entirely different memory area. Crashes that are not reproducible are likely to be caused by faulty memory chips, CPU overclocking, or overzealous CPU cache settings. The other kind of reasons---a bug in GCC---tends to be more consistent. That is, it usually crashes consistently if you reproduce the steps exactly from the system startup on. > 1) it cannot be a h/w problem because everything works when compiling > in Linux with GCC. I don't think compiling Linux is a proof. > 2) The RHIDE memory display decreases and stays at a lower value when > finished This is normal behavior. > 4) Sysmon32 always shows <10MB free memory and a >50MB disk cache. Does the amount of free memory oscillate between ~5MB and 0 during compilation, going down to 0 when cc1.exe or cc1plus.exe runs, then going up again when it exits? If so, this is normal behavior, and it means that the compiler gets enough memory to do its thing. In other words, the ICE is not triggered by the lack of memory, but by something else. > #include > int _crt0_startup_flags = _CRT0_FLAG_UNIX_SBRK; > > But I do not have the djgpp source to change this! You could download them from SimTel, introduce this line, and rebuild the compiler. FWIW, the cases where the Unixy sbrk helped in the past were not ICEs, the compiler actually crashed with SIGSEGV.