{"commits":[{"author":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T13:42:05.000000-05:00"},"committer":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T13:42:05.000000-05:00"},"extraHeaders":[{"key":"gpgsig","value":"-----BEGIN PGP SIGNATURE-----\n\niHUEABYKAB0WIQRTb8rPOazqudeTFssbuet+pgImWwUCapxifQAKCRAbuet+pgIm\nW2bxAQCQ3tM5VZWfQHuZgUh1pFnb3B1e96NRsCkxxAa73cuo0wD/buODgJZp8bmt\nSTXP3xPPfkj4ItKfpmzYyWyEMdwZZgk=\n=gcRc\n-----END PGP SIGNATURE-----\n"}],"message":"Follow as65 into a dozen more corners\n\nComparing the new debug files across every probe turned out to be a good\nlens on the rest of the assembler: most of what it found had nothing to\ndo with the format. Each of these was measured against the original\nbefore anything here changed.\n\nA label, and a label given a value with `equ`, keeps the value pass one\ngave it. Pass two works out a value and shows it in the listing, but\nwhat every reference to that label is worth was settled in pass one, and\nonly `set` is re-evaluated. That is why as65 never reports the phase\nerror it has a message for, and why the `-g` file's symbol values are\nthe ones the table held between the passes; nothing here reports that\nerror any more either.\n\nA page break still waiting when the listing ends is dropped rather than\nstarting a page with nothing on it -- unless the page it would follow\nran two lines or more past its length, which is what happens when the\nthree-line block reporting what was written goes down on a page with one\nline left. Both halves of that were measured across a matrix of page\nlengths, widths, listing lengths and closing directives.\n\nThe smaller ones, each with its own comment where it is implemented:\n\n- `org` swallows its label without complaint and without entering it,\n and its value must fit in sixteen bits, sign and all.\n- A label on a line that failed is still entered, holding the location\n counter: a bad `equ`, an unknown mnemonic, or a mnemonic that could\n not be read at all -- which is \"Invalid mnemonic\" rather than\n \"Expecting label or opcode\" once a label has been read.\n- `equ` no longer swallows whatever follows its expression.\n- `a`, `x` and `y` name registers and `hi` and `lo` are operators, so\n none of the five is a value or a label name.\n- The four brackets and anything from `del` up are \"Invalid format\"\n wherever they fall in an expression; a control character below a space\n is white space.\n- `page ` takes one to ninety-nine and refuses anything else, listing\n or no listing.\n- An unnamed `ds ` in a `struct` shows no offset in the listing.\n- The listing echoes a source line with every control character below a\n space taken out, comments included.\n- A branch whose target is out of reach is still costed in the listing,\n against the target it could not reach.\n\nThe command line was gone over at the same time. The source name is\nlooked for with `.a65`, `.asm` and `.s` pasted on the end of it when the\nname as given is not there; the file diagnostics are as65's wording and\nexit status; `-?` prints the option list and the bare command prints the\nusage line alone; `-h` with no number says so and carries on, and `-o`\nwith no name is ignored rather than opening a file called nothing.\n\nTwo things as65 does are deliberately not followed, and the README says\nso: a forward-referenced `bra` on a 6502X, which it sizes as a short\nbranch and emits as the undocumented $80, and `\\\"` inside a\ndouble-quoted string, which segfaults it.\n\nCo-Authored-By: Claude Opus 5 (1M context) \nClaude-Session: https://claude.ai/code/session_01XL9YYos6VdJAQhQJsZcJr5\n","oid":"9e9cb0b1d23751ab4acce8343feefa53d80b7786","parents":["cecaf8afadbe01846f68c45d93348e848dbcddc0"],"tree":"ac201061a20f82a627b8d8c3ffe12791f56de3c3"},{"author":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T11:11:39.000000-05:00"},"committer":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T11:11:39.000000-05:00"},"extraHeaders":[{"key":"gpgsig","value":"-----BEGIN PGP SIGNATURE-----\n\niHUEABYKAB0WIQRTb8rPOazqudeTFssbuet+pgImWwUCapw/OwAKCRAbuet+pgIm\nW589AP0d8Qzk+KRQNBJwb3ztpzzn2Y6F9Y8D5UZhxcZQr9I5zQD+LjzkZYxjkLnH\nubjjqX7oAFRitgEg1QjvQXmRPuwr1QY=\n=xpbp\n-----END PGP SIGNATURE-----\n"}],"message":"Call the original as65 rather than as65 [1.42]\n\nThe version was there to keep the original apart from ours while the two\nshared a name, and they no longer do. It stays only where the text is\nquoting what as65 actually prints: the banner in the README, in\nsrc/banner.zig's explanation of why we do not print it, and in the\nconstant the rename filter matches on.\n\nAlso drops a doc comment that had been written twice.\n\nCo-Authored-By: Claude Opus 5 (1M context) \nClaude-Session: https://claude.ai/code/session_01XL9YYos6VdJAQhQJsZcJr5\n","oid":"b7beab242512e2fe3595eaed953e1de296022c71","parents":["dbed412ac7134289d55978c1bcb5a97f42f3e062"],"tree":"0482c33d5b2fb33b35493f07aab8f694b68deb8d"},{"author":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T09:55:14.000000-05:00"},"committer":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T09:55:14.000000-05:00"},"extraHeaders":[{"key":"gpgsig","value":"-----BEGIN PGP SIGNATURE-----\n\niHUEABYKAB0WIQRTb8rPOazqudeTFssbuet+pgImWwUCapwtUgAKCRAbuet+pgIm\nW/vSAP0QFCeaES3ShC+wEAyq0DjyafXqoOk/dpBQyI+KnWh+SwEAvoW/0aFJrjNZ\nAYg0LHjBXurBTgS6a+gp8jzHmGl0Qw8=\n=b6Ic\n-----END PGP SIGNATURE-----\n"}],"message":"Match as65's listing pagination and error reporting\n\nThe listing side of as65 turns out to have three different rules for\nwhen a page breaks, and which one applies depends on what is being\nwritten. Ordinary listing lines break at the page length; the symbol\ntable's entries do too, but the three lines that close it are never\nsplit from them, and the summaries are allowed a line past the bottom\nbefore they break. The block that reports what was written is looked at\nonce and then written straight through, which is how a listing at -h60\ncomes to hold sixty-two lines. A page length below twenty is raised to\ntwenty, and `page ` breaks only when the lines asked for would run\nmore than four past the end.\n\nAlso measured and reproduced: `end` records a start address without\nstopping the assembly, and shows the last value the listing displayed\nrather than the address it was given; a label is not entered in the\ntable until its own line has been assembled, so `Target jsr Target`\nassembles in pass one as a jump to zero; `org` onto a byte already\nassembled is refused; an `equ` whose expression will not evaluate still\nleaves the label behind holding the location counter; and a label may\ncontain digits but not begin with one.\n\nAcross a matrix of 480 option combinations on as65's own test source,\n458 now agree byte for byte; the rest are pagination corners that need\nboth a short page and a symbol table or a pass 1 listing.\n\nCo-Authored-By: Claude Opus 5 (1M context) \nClaude-Session: https://claude.ai/code/session_01XL9YYos6VdJAQhQJsZcJr5\n","oid":"93426f0aebb172f69ab5fff8945c63741ba33558","parents":["b5535d49b845f482e5d6a233d926ee6b28da2b42"],"tree":"a24f228357b8d2d673e1206abf5f6535f8c139b9"},{"author":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T09:31:01.000000-05:00"},"committer":{"email":"jeff@ocjtech.us","name":"Jeffrey C. Ollie","when":"2026-09-05T09:31:01.000000-05:00"},"extraHeaders":[{"key":"gpgsig","value":"-----BEGIN PGP SIGNATURE-----\n\niHUEABYKAB0WIQRTb8rPOazqudeTFssbuet+pgImWwUCapwnpQAKCRAbuet+pgIm\nW6s0AQCBHcnNJym3BKZMX98X9TjMz3ft2ZLPDOv6z9ILCLh37QD/WSmOQvjnEpi+\n3yj7l6btOnj3O54hduP8mJfip/uOzwM=\n=JogA\n-----END PGP SIGNATURE-----\n"}],"message":"Add a 6502 assembler in the as65 dialect\n\nAn assembler for the 6502 and 65SC02 that reproduces Frank Kingswood's\nas65 [1.42] byte for byte: the binary, the s-records, the intel-hex and\nthe listing, column for column.\n\nas65 is MIT licensed but its source was never published, so the tables\nand the behaviour here were measured from the binary rather than read\nout of it. Every opcode, cycle count and addressing-mode rejection came\nfrom assembling one mnemonic and mode at a time and reading the listing\nback; the odd rows that no data sheet would agree with -- sty ,x\nassembling to stz , jsr () dropping its parentheses, bit #imm\naccepted on a plain 6502 -- are reproduced deliberately and commented\nwhere they appear.\n\nThe same goes for align, whose fill count as65 works out as\n(boundary - pc) in unsigned 32-bit arithmetic, so that past the boundary\nthe subtraction wraps and the result lands congruent to 2^32 mod\nboundary; and for the symbol table's collation, which is neither ASCII\norder nor any locale's.\n\nas65 -x -q -h0 on its own test source now agrees exactly, with and\nwithout -c, -t, -p, -m, -s, -s2, -n, -i, -z, -x1 and -h.\n\nCo-Authored-By: Claude Opus 5 (1M context) \nClaude-Session: https://claude.ai/code/session_01XL9YYos6VdJAQhQJsZcJr5\n","oid":"b5535d49b845f482e5d6a233d926ee6b28da2b42","parents":["75e200e9152c0299597ef84e1cb2da8a60b3bb77"],"tree":"8de72e93d04762bf91e3cf67920cf0e34f8b25e8"}]}

Version history of src/expr.zig

September 2026