“It Can’t Be a Bug, My Makefile Depends on It!” 195 free until they are put into extraordinary circumstances, whereupon they mysteriously fail. Lisp implementations usually have real exception-handling systems. The exceptional conditions have names like OUT-OF-MEMORY and the pro- grammer can establish exception handlers for specific types of conditions. These handlers get called automatically when the exceptions are raised— no intervention or special tests are needed on the part of the programmer. When used properly, these handlers lead to more robust software. The programming language CLU also has exception-handling support embedded into the language. Every function definition also has a list of exceptional conditions that could be signaled by that function. Explicit lin- guistic support for exceptions allows the compiler to grumble when excep- tions are not handled. CLU programs tend to be quite robust since CLU programmers spend time thinking about exception-handling in order to get the compiler to shut up. C programs, on the other hand… Date: 16 Dec 88 16:12:13 GMT Subject: Re: GNU Emacs From: debra@alice.UUCP In article 448@myab.se lars@myab.se (Lars Pensj) writes: ...It is of vital importance that all programs on their own check results of system calls (like write).... I agree, but unfortunately very few programs actually do this for read and write. It is very common in Unix utilities to check the result of the open system call and then just assume that writing and closing will go well. Reasons are obvious: programmers are a bit lazy, and the programs become smaller and faster if you don’t check. (So not checking also makes your system look better in benchmarks that use standard utili- ties...) The author goes on to state that, since most Unix utilities don’t check the return codes from write() system calls, it is vitally important for system administrators to make sure that there is free space on all file systems at all time. And it’s true: most Unix programs assume that if they can open a file for writing, they can probably write as many bytes as they need. Things like this should make you go “hmmm.” A really frightening thing about the Miller et al. article “An Empirical Study of the Reliability of Unix Utilities” is that the article immediately preceding it tells about how

196 Programming Mission Control at the Johnson Space Center in Houston is switching to Unix systems for real-time data acquisition. Hmmm. Catching Bugs Is Socially Unacceptable Not checking for and not reporting bugs makes a manufacturer’s machine seem more robust and powerful than it actually is. More importantly, if Unix machines reported every error and malfunction, no one would buy them! This is a real phenomenon. Date: Thu, 11 Jan 90 09:07:05 PST From: Daniel Weise daniel@mojave.stanford.edu To: UNIX-HATERS Subject: Now, isn’t that clear? Due to HP engineering, my HP Unix boxes REPORT errors on the net that they see that affect them. These HPs live on the same net as SUN, MIPS, and DEC workstations. Very often we will have a prob- lem because of another machine, but when we inform the owner of the other machine (who, because his machine throws away error messages, doesn’t know his machine is hosed and spending half its time retransmitting packets), he will claim the problem is at our end because our machine is reporting the problem! In the Unix world the messenger is shot.