nullglob is documented incorrectly
Today, a user in the #bash IRC channel of libera.chat, misunderstood how nullglob is supposed to work because is documented incorrectly in the man page: 'nullglob' If set, Bash allows filename patterns which match no files to expand to a null string, rather than themselves. globs expand to nothing if there are no matches with nullglob set, not to the null/empty string. o/ emanuele6
Re: nullglob is documented incorrectly
On Sun, Nov 05, 2023 at 11:15:50AM +0100, Emanuele Torre wrote: > Today, a user in the #bash IRC channel of libera.chat, misunderstood how > nullglob is supposed to work because is documented incorrectly in the > man page: > > 'nullglob' > If set, Bash allows filename patterns which match no files to > expand to a null string, rather than themselves. > > globs expand to nothing if there are no matches with nullglob set, not > to the null/empty string. > > o/ > emanuele6 It may even be better to say that the patterns that do not match any names are *removed*. That way one avoids the complications and ambiguities of explaining what the pattern expands or not expands to. -- Andreas (Kusalananda) Kähäri Uppsala, Sweden .
[PATCH] fdflags: fix -s with multiple flags, and multiple fd arguments
fdflags -s was re-parsing the setspec for each fd argument. This is problematic because it is using strtok() to parse it destructively. So, only the first part of the setspec will be set for all file descriptors; the other parts will only be set for the first file descriptor: $ enable fdflags $ exec {foo1} /dev/null $ fdflags -- "$foo1" "$foo2" 10:nonblock,cloexec 11:nonblock $ exec {foo1}<&- {foo2}<&- {foo3}next) { @@ -335,7 +336,7 @@ fdflags_builtin (WORD_LIST *list) } num = inum; /* truncate to int */ if (setflag) - setone (num, setspec, verbose); + setone (num, setpos, setneg, verbose); else printone (num, 1, verbose); } -- 2.42.0