diff --git a/general-info/release-20c.txt b/general-info/release-20c.txt index 46ef7575f56d7e7f4098a391925819649fc44867..424f1d0691a3550e1ac4a1b60920c7efa5318b9a 100644 --- a/general-info/release-20c.txt +++ b/general-info/release-20c.txt @@ -33,9 +33,27 @@ New in this release: errors are signaled on invalid sequences instead of silently replacing the bad sequence with some kind of replacement character. + - In 19f through 20b, READ-SEQUENCE could read mostly arbitrary + data from a stream in to the given sequence. In this release, + READ-SEQUENCE cannot do that unless the stream is a + binary-text-stream. This is an incompatible change from + previous releases. * ANSI compliance fixes: - + - Fixes for signaling errors with READ-CHAR and READ-BYTE + o READ-CHAR signals errors if the stream is not a character + stream. This is a change from 20a and 20b, but matches + releases before 19f. (Almost. 19f allowed reading characters + from (unsigned-byte 8) streams. 19e did not.) + o READ-BYTE signals errors if the stream is not a binary + stream. This is also a change from 20a and 20b, but matches + releases before 19f. (Almost. 19f allowed reading bytes from + character streams. 19e did not.) + o But READ-CHAR and READ-BYTE will work if the stream class is + 'binary-text-stream, an extension for bivalent streams in + CMUCL. READ-CHAR will use the specified external format for + such streams. READ-BYTE reads (unsigned-byte 8) elements from + such streams. * Bugfixes: - The pairwise composition table is now correctly built. Previously, it skipped over non-BMP codepoints. This also