Re: why fail , [[ $sep$*$sep == $sep--$sep ]]

[Kerin Millar]

On Sun, 18 Feb 2024 15:56:45 -0500
Lawrence Velázquez <vq@larryv.me> wrote:

> On Sun, Feb 18, 2024, at 8:51 AM, Kerin Millar wrote:
> > On Sun, 18 Feb 2024 14:34:43 +0100
> > alex xmb sw ratchev <fxmbsw7@gmail.com> wrote:
> >
> >> On Sun, Feb 18, 2024, 14:18 Kerin Millar <kfm@plushkava.net> wrote:
> >> 
> >> > On Sat, 17 Feb 2024 17:01:13 +0100
> >> > alex xmb sw ratchev <fxmbsw7@gmail.com> wrote:
> >> >
> >> > > oh i know why it failed
> >> > > set interprets the first -- away
> >> >
> >> > That is not why it is failing to behave as you expect. Rather, the issue
> >> > is that your conditional expression is incorrect. It should be [[ 
> >> > "$f$*$f"
> >> > == *"$f--$f"* ]]. Note that you have chosen to change the value of IFS. 
> >> > It
> >> > is important that you do not omit the quotes - not even for the left hand
> >> > side of the expression!
> >> >
> >> 
> >> i thought there is a ' flat left side of == '
> >
> > Unfortunately, the raw output of xtrace is not much use for human 
> > eyeballs. Here's the exact difference.
> >
> > $ ( set aa bb cc -- dd; f=$'\1' IFS=$f PS4= BASH_XTRACEFD=1; set -x; [[ 
> > "$f$*$f" == *"$f--$f"* ]] ) | hexdump -C
> > 00000000  5b 5b 20 01 61 61 01 62  62 01 63 63 01 2d 2d 01  |[[ 
> > .aa.bb.cc.--.|
> > 00000010  64 64 01 20 3d 3d 20 2a  01 5c 2d 5c 2d 01 2a 20  |dd. == 
> > *.\-\-.* |
> > 00000020  5d 5d 0a                                          |]].|
> >
> > $ ( set aa bb cc -- dd; f=$'\1' IFS=$f PS4= BASH_XTRACEFD=1; set -x; [[ 
> > $f$*$f == *"$f--$f"* ]] ) | hexdump -C
> > 00000000  5b 5b 20 61 61 62 62 63  63 2d 2d 64 64 20 3d 3d  |[[ 
> > aabbcc--dd ==|
> > 00000010  20 2a 01 5c 2d 5c 2d 01  2a 20 5d 5d 0a           | *.\-\-.* 
> > ]].|
> 
> This surprises me because IFS is not lost if it's set to something
> a little less ... weird:
> 
> $ (set aa bb cc -- dd; f=+ IFS=$f; set -x; [[ $f$*$f == *"$f--$f"* ]])
> + [[ +aa+bb+cc+--+dd+ == *\+\-\-\+* ]]
> 
> Why the difference?

Why indeed. This must surely be a bug. Neither of + and $'\1' are the standard 
whitespace characters (in view of field splitting). There is no reason for them 
to be handled inconsistently.

-- 
Kerin Millar