Interesting features of the GNU assembler Gas

Interesting features of gas:

• multiple subsections: .data 0, .data 1, etc. You can switch around between them during assembly, and then they get concatenated when assembly is done. There’s some ambiguity in the manual about how this interacts with the “.” symbol.
• The “.” symbol, which is like Intel’s $, and which can be reassigned to move around and assemble things in funny places. I haven’t yet tried using it to assemble code someplace we’ve already been (e.g. for a counted string). The documentation for .org says this won't work.
• .fill would be useful for filling an area with a pattern (despite its bizarre semantics.)
• there is of course a full conditional compilation system.
• there is of course a full macro facility, with default arguments, named arguments, recursive macros, string interpolation, the ability to redefine existing symbols, and optionally local variables.
• .incbin lets you suck in an external file, or part of an external file, as data.
• It uses Knuth's 1f, 1b, 2f, 2b local labels. These work remarkably well for something so archaic-looking.
• The -a option generates an assembly listing. There's a bunch of formatting stuff for the listings: .eject, .title, .sbttl, .nolist, .list, .psize.
• It supports bignum math and emitting 8- and 16-byte integers from them.
• There’s a “section stack” so you can assemble into a different section for a while, then pop back to where you were, or even swap back and forth between two sections without knowing which ones they are. Unfortunately, at the moment, the section stack is only partly implemented as documented.
• .print prints stuff out.
• ELF apparently has a bunch of section flags: “allocatable”, writable, executable, "mergeable", and a few more
• there’s sort of support for defining struct fields with .struct
• .intel_syntax supports Intel assembler syntax.