With the new experimental updates I have added a couple of directives that assist in handling a wider array of syntaxes with fasmg.

By default, fasmg's preprocessor strips comments (the ones that start with semicolon) from the lines before allowing them to be processed further. Therefore it was impossible to parse with macros any syntax that would use semicolons for a different purpose.

The new RETAINCOMMENTS directive switches preprocessor into mode where it no longer does that. It is then possible to use semicolon as any other special characters, and process them with MATCH to handle any syntax that need to interpret them:

The preprocessor of fasmg also combines the lines when the line break is preceded by a backslash (just like preprocessor of fasm 1 did) and also might sometimes be a obstacle when parsing a specific syntax. So I also added a ISOLATELINES directive that disables it. You can use these two directives combined to switch fasmg's preprocessor into mode that allows to process just raw tokenized text.

What kind of complications might one encounter when trying to wrap lines in ISOLATELINES / COMBINELINES or RETAINCOMMENTS / REMOVECOMMENTS? For example, it wouldn't be sufficient to wrap the above macro - as the preprocessing of following lines would strip comments obviously. It seems a great deal of care is needed in a larger context to get the desired result.

Nonetheless a very powerful feature.

I've experimented with here: https://board.flatassembler.net/topic.php?p=208437#208437

Yes, of course, once you start switching between the modes it may become very tricky.

One thing to consider is that these preprocessing options only apply to lines read from source text, lines made by macros are not processed in such way (as they were normally not expected to contain comments, for example). Therefore once you define a macro that contains non-stripped semicolons, they are going to stay there even if you call this macro while having REMOVECOMMENTS mode enabled.

Note that this applies to not just macros, but any symbolic values. Once a semicolon (or backslash at the end of line) gets "smuggled" into a value somewhere, it is no longer going to be stripped even with REMOVECOMMENTS option brought back. The option only applies to processing new lines read from source.