mike_lee_mehr@cup.portal.com (01/10/88)
In line with some previous articles about MSCv5.0 bugs, I thought I'd relate some experiences we have had at work with the beast. We are compiling a large system of code that uses overlays and needs PLINK-86 to link it up, so the chances of bugs creeping in are even greater. We noticed that notorious stack space problem early on, and a call to MS Customer Support turned up the data-sharing feature of the stack segment. So we determined to use the /Gt option to put all the data we could into separate segments. This works fine on MSCv4.0 but causes v5.0 to create about 32K of extra initialized data, which is what we were trying to eliminate in the first place! Further calls to MS netted the result that, while they realize this is a bug, the fix to QuickC will be out in about a month, but they "have no idea when" a bug fix to Cv5.0 will be forthcoming. So, a word of warning, added to all the others. Don't use /Gt:1 on v5.0. Also, perhaps one should think twice about using v5.0 at all until the bug fixes have begun to be circulated. Not that this disillusionment isn't common, but to be fair to Microsoft, they are trying. Patience may yet pay off. For now I think we will be sticking with v4.0, although we may use some of the new library functions if this works out.