+ lijjrRt^RIHt^RIHt!RR]4t]!RR44tR RRlltR R lt R #) aIntel HEX encode/decode. Intel HEX itself is a public, well-documented format (unlike the Roboteq bytecode it will eventually carry -- see docs/reverse_engineering.md), so this module is implemented for real rather than stubbed. It's also useful right now for reverse-engineering work: `read_intel_hex` can load .hex files exported from RoborunPlus so they can be diffed byte-for-byte. Only record types 00 (Data) and 01 (End Of File) are handled -- Roboteq scripts are small enough that extended segment/linear address records (02/03/04/05) are not expected to be needed, but that assumption should be verified against real exported files during reverse engineering. ) annotations) dataclassc]tRt^tRtR#)HexFormatErrorN)__name__ __module__ __qualname____firstlineno____static_attributes__rm\\thalassa.shore.mbari.org\ProjectLibrary\ESP\FIDO\PumpController\compiler\src\microbasic_compiler\hexfile.pyrrsr rcX]tRt^t$R]R&R]R&R]R&R]R&RRltR R ltR tR #) HexRecordint byte_countaddress record_typebytesdatacV^8dQhRR/#)returnrr)formats"r __annotate__HexRecord.__annotate__s#r c VPVP^, ,VP^,,VP,\VP4,pV)^,#))rrrsumr)selftotals& r checksumHexRecord.checksums` OO||q  "||d" $ $))n   $r cV^8dQhRR/#)rrstrr)rs"r rr)sCCCr c &\VPVP^, VP^,VP.4VP,pRVP 4P 4,VP4R ,#)r:02X)rrrrrhexupperr!)rbodys& r to_lineHexRecord.to_line)sgdoot||q'8$,,:MtO_O_`adhdmdmmTXXZ%%''T]]_S,ABBr rN)rrr r __annotations__r!r+r rr r rrs(O L KCCr rc(V^8dQhRRRRRRRR/#)rrr start_addressrbytes_per_linerr$r)rs"r rr.s)##%###VY#r cV^8gV^8d \R4h\V4V,pV'dVRW#, ,,p.pTp\^\V4V4F]pWWb,p\\V4VR,^VR7pVP VP 44V\V4, pK_ \^^^RR7p VP V P 44RP V4R,#)aEncode `data` as Intel HEX text (data records + trailing EOF record). CONFIRMED (gosub_distance corpus, 2026-07-28): real RoborunPlus-exported .hex always pads the final line out to a full `bytes_per_line` (16) bytes with 0xFF, rather than emitting a short last record for whatever remainder doesn't divide evenly. Every one of the 6 distance test cases (12/212/262/267/512/962 raw bytes) came back from RoborunPlus padded up to the next 16-byte boundary (16/224/272/272/... bytes) with the difference being solely trailing 0xFF -- 0 differing bytes in the shared range in all 6 cases, confirming GoSub codegen itself is correct (including across the 256-byte 1-vs-2-byte-address boundary). Padding here so our .hex output matches real compiler output byte-for-byte. z(bytes_per_line must be between 1 and 255irrrrr  ) ValueErrorlenrangerappendr+join) rr/r0 remainderlinesaddroffsetchunkrecordeofs &&& r write_intel_hexrA.s^c1CDDD N*Ig!;<<E D3t9n5f56c%j$-UY`ef V^^%& E  6 q!C HC LL 99U d ""r c V^8dQhRRRR/#)rtextr$rrr)rs"r rrPs""""r c\4p\VP4^R7EF5wr#VP4pV'gK VP R4'g\ RV R24h\ PVR,4p\T4^8d\ RT R24hT^,T^,T^,T^,3wrxrT^^T,p T^T,,p \Yx^,T ,YR 7p T P4T 8wd\ RT R 24hT ^8XdTPT 4EKT ^8XdM\ RT R T R 24h \ V4# \dp\ RT R24ThRp?ii;i) zDecode Intel HEX text back into a flat byte string (data records only, concatenated in file order -- does not attempt to place bytes at their absolute addresses since that's not needed for diffing bytecode dumps). )startr&zline z: missing ':' start code:NNz: invalid hex dataNz: record too shortr3z: checksum mismatchz: unsupported record type 0xr') bytearray enumerate splitlinesstrip startswithrrfromhexr5r6rr!extend)rCoutlinenoraw_lineliner*excraddr_hiaddr_lorpayloadr!r?s& r read_intel_hexrVPs +C%doo&7qAA~~ s## 50H!IJ J N==b*D t9q= 50B!CD D48GT!Wd1gtTUw4V1 Wq1z>*J'ja<7:R`kz ??  ( 50C!DE E $  JJw  D  50L[Y\L]!^_ _5B8 :) N 50B!CD# M Ns-E++ F 6FF N)) __doc__ __future__r dataclassesr ExceptionrrrArVrr r r]sE #! Y  CC C*#D"r