Version 1.4c published

[EkBass]

News and changes

These changes are about the current source code. These are effected once new binary release is published

4th September 2026

Published as version 1.4c

Bug fix: ROUND() used banker's rounding instead of always rounding halves up

ROUND() called Math.Round() without a midpoint mode, so .NET defaulted to MidpointRounding.ToEven ("banker's rounding"): ROUND(0.5) gave 0, ROUND(10.5) gave 10, ROUND(2.5) gave 2 — always rounding .5 to the nearest even number instead of always up. Now uses MidpointRounding.AwayFromZero, so .5 always rounds away from zero:

PRINT ROUND(0.5)    ' now 1 (was 0)
PRINT ROUND(10.5)   ' now 11 (was 10)
PRINT ROUND(-0.5)   ' now -1

Bug fix: SRAND() had a duplicate '0' and a stray '£' in its character set

SRAND(n)'s allowed character set accidentally listed 0 twice (giving it double the chance of the other digits) and included the British pound sign £, a non-ASCII character that had no business being there. The set is now 81 characters, each with equal probability, £ removed entirely.

Bug fix: PRG_ROOT# was missing its trailing directory separator

PRG_ROOT# never ended with \ or /, so concatenating a relative path directly (PRG_ROOT# + "data/file.txt") merged the program's folder name with the subfolder name into one broken path (e.g. ...\MyProjectdata\file.txt) and silently failed to create the file. PRG_ROOT# now always ends with a directory separator:

FILEWRITE PRG_ROOT# + "data/file.txt", content$   ' now works without adding your own separator

Bug fix: FILEREAD-into-array (.env style parsing) broke completely on a single comment line

Assigning FILEREAD()'s output to a DIM'd array (DIM env$ = FILEREAD(".env")) parses KEY=value lines into array elements. The parser required every line to contain =, so a single #-prefixed comment line (no = at all) made the whole population step silently fail — leaving the array completely empty, including keys that appeared before the comment. Blank lines and full-line # comments are now skipped before validation. Inline comments (KEY=value # note) are not yet supported.

Bug fix: calling a DEF FN function without the FN keyword silently did nothing

Writing Foo$(1) on its own line (forgetting the FN keyword, e.g. instead of FN Foo$(1)) was silently swallowed — the parser treated it as an array-element assignment target, found no = sign after it, and just moved on without calling the function or reporting anything. Now, if the name matches a declared DEF FN function, this produces a clear error instead:

DEF FN Foo$(bar$)
    RETURN bar$ + 1
END DEF

Foo$(1)   ' Error: 'FOO$' is a DEF FN function, not an array — did you forget the FN keyword? Use: FN FOO$(...)

Examples folder reorganized

Examples/ was a flat pile of ~80 scripts. Split into Examples/graphics/, Examples/audio/, Examples/console/ and Examples/classics/ (ports of classic 1970s/80s BASIC games — Eliza, Acey Ducey, Robots, Towers of Hanoi, Trap, Russian Roulette, Rock-Paper-Scissors, the Wingbermuehle maze generator). Also removed 12 old internal regression-test scripts (*_test.bas style files exercising ISSET, FSTRING, line continuation, FastTrig, DISTANCE, HTTP listen/timeout, etc.) that had no demonstration value and were already covered by the automated test pass long ago. rosetta-code/ is untouched and stays a separate, curated category.

Going forward, more effort should go into graphics examples specifically. BazzBasic still gets mistaken for a console-only interpreter by a lot of people, partly because so many of the earlier example programs were console-based — the graphics/SDL2 side deserves better representation.