From mboxrd@z Thu Jan 1 00:00:00 1970 From: mrs@wrs.com (Mike Stump) To: egcs@cygnus.com, schmid@ltoi.iap.physik.th-darmstadt.de Subject: Re: testsuite results Date: Mon, 13 Oct 1997 13:41:00 -0000 Message-id: <199710132015.NAA04846@kankakee.wrs.com> X-SW-Source: 1997-10/msg00505.html > Date: Sun, 12 Oct 1997 19:44:51 +0200 (MEST) > From: Peter Schmid > To: egcs@cygnus.com > 2. While compiling g++.mike/ambig1.C I get an recoverable compiler error. > What is that? Well, the general rule is to fix the first problem listed, and then recompile. In this case, that is what should be done, though, since this testcase just wants to generate error messages, of course, there is no problem. If you want a more complete description, imagine a compiler that could detect when it was in error, then corrected it's problem before screwing the user, and found a way to get the developers to fix the underlying problem, to further enhance the compiler... Well, that is what it does. In this case, since there was a prior error, it is less meaningful, but in theory we could teach the compiler to recover from the first error in a more complete way. Currently, it doesn't recover from the first error as well as it should. One might be tempted to call this software fault tolerance, personally I think it is more useful than core dumping the compiler.