After a bit of fiddling around with what I thought would be a lexical analyzer for Flex, I now realize that the lexer isn’t defined by scan.l, but a combination of one generated from flex.skl, a file that is processed by m4, and the scanner generated by scan.l. Originally, flex.skl included some switches for C or C++, but it’s now m4 dependent. I’m not sure why #ifdefs wouldn’t have sufficed; it would have been far more portable. Further, after carefully inspecting the parser grammar, I now realize it is missing a ton of things from an input file because the lexer yanks so much information out of the input stream and never passes it onto the parser. For example, most of the options in a Flex file are simply swallowed by the scanner, only the ones with ‘=’ passed to the parser. (Bullshit.) Code blocks within %{ … %} or %top{ … } are swallowed up by the lexer, and never mentioned in the parser grammar. So, in order to really write an import of Flex, the lexer and scanner would have to be written from scratch. (Originally, I was toying with doing just that, after successfully converting Bison’s parse.y to Antlr, but chose to start with the Flex sources because I thought the grammars were complete. They just were complete garbage from the standpoint of an import.) Given the number of problems, I’m going to pass on Flex import for now. I will probably revisit Flex import but at a later date. For now, I’m ODed on a fruitless translation.
Fortunately, I don’t really need to implement any of this for Antlrvsix right now. I’m going to continue with the rest of the “to do’s” I have slated.