Skip to content

GoldED+ 2.0.0-20260912

Pre-release
Pre-release

Choose a tag to compare

@evs38 evs38 released this 11 Sep 22:03
· 4 commits to unicode since this release

GoldED+ 2.0.0, the first snapshot under that number: the unicode
branch has been the working editor of its users for a month, and the
version says so. What changed since 20260905.

The header no longer flashes garbage while a message is saved. The
From, To and Subject fields were converted to the message's charset in
place before the header was painted; a Russian subject saved into a
CP866 echo from a UTF-8 session showed as a row of � for a moment. The
fields stay in the session's charset for as long as the message lives
in memory and are converted only as the message base takes them. This
also stops the address book from learning names in CP866, and keeps a
confirmation receipt's header readable while its AKA picker is up.

The Windows clipboard. In a UTF-8 session GoldED+ pasted its own
copy as three characters per letter: the 8-bit clipboard formats held
the raw UTF-8 bytes and the paste side picked the first of them. Every
format now holds the text, and paste prefers Unicode.

XLATUTFAUTODETECT MIXED, experimental: mail that quotes UTF-8 into
a CP866 reply and declares CP866 for the lot - subject and quotes in
one charset, the author's own lines in another - is read paragraph by
paragraph and field by field. Includes OVERRIDE. Measured on a base of
98 000 messages: none of a Russian echo's paragraphs changed charset by
mistake once the CJK ideographs were left out of the per-line test
("фев" in a quote header had passed for one).

A terminal narrower than eighty columns - a phone with a large font

  • crashed the reader below the header's node column and made the
    message list exit; both were arithmetic that assumed the screen it was
    laid out for. Checked down to eight columns under AddressSanitizer.

WRITEUCSHEADERS measures against the FTS-0001 widths, 35, 35 and
71 bytes, whatever LARGEHEADERTOBASE lets the base keep: the kludge is
for the reader at the far end of the packet. With the base's widths a
JAM area wrote no kludge under a hundred bytes and a Synchronet area
none at all.

A kludge line longer than the screen left the line after it unread
as a kludge; FMail writes UCSSUBJ before CHRS, so with a long subject
the message came out in the area's charset instead of its own - on the
one terminal where the line wrapped. Fixed, and the charset kludge is
now looked for before the kludge lines ahead of it are converted.

Area separators on a CP866 Windows console drew as a row of
question marks: a box-drawing byte went into the console cell as a
codepoint. Every run of such characters - scrollbars, window titles -
goes through the session's charset now.

FGHI URLs. area://ECHO?msgid=2:5020/1+12345678 in a message is
recognised as a URL and, chosen from the URL list, followed inside
GoldED+: the echo is opened and the message with that MSGID shown.
DISPHDRFGHIURL <no/yes/short> puts the message's own URL under the
header, and the template token @ofghiurl gives the quoted message's.
Ported from Konstantin Kuzov's GoldED-NSF patch of 2009.

Seventeen archives. Every binary carries the build stamp 20260912.