Issue #21625 has been updated by nobu (Nobuyoshi Nakada). https://github.com/ruby/ruby/pull/14985 ---------------------------------------- Bug #21625: Allow IO#wait_readable together with IO#ungetc even in text mode https://bugs.ruby-lang.org/issues/21625#change-114970 * Author: YO4 (Yoshinao Muramatsu) * Status: Open * Backport: 3.2: REQUIRED, 3.3: REQUIRED, 3.4: REQUIRED ---------------------------------------- ### Background IO#wait_readable calles rb_io_check_byte_readable() in that C implementation, so that is treated as binary reading class operation. ```ruby open(IO::NULL, "rt") { |f| f.ungetc(?a); f.wait_readable } # => in 'IO#wait_readable': byte oriented read for character buffered IO (IOError) ``` This is inconsistent with the behavior of IO#select. ```ruby open(IO::NULL, "rt") { |f| f.ungetc(?a); IO.select([f]) } # => [[#<File:NUL (closed)>], [], []] ``` ### Proposal Change IO#wait_readable into character reading class function. This allows mixing usage of IO#wait_readable and IO#ungetc. This change avoids the IOError that previously occurred, but in that case, an exception is thrown when the binary read method is subsequently called. Overall behavior remains unchanged, but exceptions will occur at more desirable locations. If the method being called next is a character method, no exception will be thrown. But that is the desired behavior. ### related methods IO#io_wait_priority may apply similar changes for consistency. IO#ready? seems to have the same issue. This is in the io-wait default gem. IO#wait already respects the character read buffer. -- https://bugs.ruby-lang.org/

Issue #21625 has been updated by alanwu (Alan Wu). Backport changed from 3.2: REQUIRED, 3.3: DONE, 3.4: REQUIRED to 3.2: REQUIRED, 3.3: DONE, 3.4: DONE ruby_3_4 commit:a7eba5453934e778c5911a4b69d9c3b8fdf31a23 ---------------------------------------- Bug #21625: Allow IO#wait_readable together with IO#ungetc even in text mode https://bugs.ruby-lang.org/issues/21625#change-115114 * Author: YO4 (Yoshinao Muramatsu) * Status: Closed * Backport: 3.2: REQUIRED, 3.3: DONE, 3.4: DONE ---------------------------------------- ### Background IO#wait_readable calles rb_io_check_byte_readable() in that C implementation, so that is treated as binary reading class operation. ```ruby open(IO::NULL, "rt") { |f| f.ungetc(?a); f.wait_readable } # => in 'IO#wait_readable': byte oriented read for character buffered IO (IOError) ``` This is inconsistent with the behavior of IO#select. ```ruby open(IO::NULL, "rt") { |f| f.ungetc(?a); IO.select([f]) } # => [[#<File:NUL (closed)>], [], []] ``` ### Proposal Change IO#wait_readable into character reading class function. This allows mixing usage of IO#wait_readable and IO#ungetc. This change avoids the IOError that previously occurred, but in that case, an exception is thrown when the binary read method is subsequently called. Overall behavior remains unchanged, but exceptions will occur at more desirable locations. If the method being called next is a character method, no exception will be thrown. But that is the desired behavior. ### related methods IO#io_wait_priority may apply similar changes for consistency. IO#ready? seems to have the same issue. This is in the io-wait default gem. IO#wait already respects the character read buffer. -- https://bugs.ruby-lang.org/