jeffe@omews63.intel.com (Jeff Eaton) (03/20/90)
I haven't seen anything on this here, so I thought I would mention it and perhaps spare somebody some grief. My environment is a Sun386, running SunOS 4.0.2. I've known for some time that certain features of X will make the server toss its cookies, in particular various uses of wide lines, but I'd never had the time to sit down and try to define the problem well. I made the time this weekend. I also made an attempt to find the problem in the server, and I made the astounding discovery that all the problems disappeared when the server was compiled -g. That's right, the sun compiler is crap. I should of remembered. In short, if you want to do X on a Sun386, use GCC (I used 1.37.1) to build the server. Or, you can not use -O. Either will build a server that I haven't been able to crash yet. While this is not a bug report for X, it would be nice if the X people could include the Sun386 in the list of recomendations for using GCC (along with 68k, etc.) Jeff Eaton Intel Corporation jeffe@mipon2.intel.com
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (03/20/90)
Your "facts" in no way substantiate your claim. Based on your evidence, it's just as likely there's an uninitialized variable somewhere that's causing the problem, and it just happens to get tickled by the particular stack/whatever environment under the 386/cc combination. Compiling with -g or with a different compiler can change that environment. Without more proof, I'd say the server code is just as suspect as the compiler. As usual, a stack trace of the crash is often necessary for diagnosis.