<p>My compiler spoke only an approximation to standard Forth. The structure of my input programs probably wouldn&#39;t have been approved by the language standard, in particular. Source files were required to begin with all the function and variable definitions, and then the main program would follow them.</p><p>So the compiler noticed when the first thing appeared in the source file that was not a function or variable definition, or nested inside one. And when it saw it, that was where it set the entry-point address of the output executable.</p><p>But the check for the first top-level thing in the source file happened before, rather than after, discarding comments. So if you wrote a comment before a function definition, the program&#39;s entry point address would point at the start of that function, instead of the start of the main program.</p><p>I&#39;d never noticed at school, because although I did write comments in my function definitions, I wrote them just _inside_ the function, where they didn&#39;t trigger this bug. But 20 years later, I dug out this compiler for nostalgia&#39;s sake, ran it in BeebEm, and tried writing a program in it (world&#39;s slowest MD5 implementation!) to demonstrate that it still worked.</p><p>And since my schooldays, I had switched to the habit that most people seem to have (and that tools like rustdoc outright enforce), of putting a function&#39;s comment _outside_ the function definition. Bang! Bad output code.</p><p>(The worst of it was that then of course pride required me to fix it, and unlike my Forth _programs_, the Forth _compiler_ was in dense uncommented 6502 assembler. Got there in the end, but ugh, the things we had to do to fit useful amounts of stuff into such small machines.)</p><p>(2/2)</p>