bashcompinit incorrectly calculates COMP_LINE (proposed fix included)

bashcompinit incorrectly calculates COMP_LINE (proposed fix included)

[Evan Underscore]

[-- Attachment #1: Type: text/plain, Size: 962 bytes --]

Hi all,

A user discovered a bug in bashcompinit that causes COMP_LINE to be
incorrectly calculated. A summary of the incorrect behaviour is here:
https://github.com/kislyuk/argcomplete/issues/228#issuecomment-327696251
(Note some of the surrounding discussion is incorrect - this bug is
unrelated to unicode.) The table at the bottom of this comment summarises
how bashcompinit differs in its calculation of COMP_POINT.

The change that caused the bug appears to have been introduced here:
https://github.com/zsh-users/zsh/commit/23907bb840c80eef99eabba17e086e44c9b2d3fc#diff-b8d49b00032bcbfcdc709649f5d68505R11

Reverting this line (i.e. using CURRENT-1 instead of CURRENT, as it was
before this commit) fixes the bug for me, causing zsh to report 18 and 19
for the two test lines respectively, matching Bash.

I am not a subscriber; please send responses to this email address. I wish
to note the resolution of this on the argcomplete ticket.

Thanks,

Evan.

Re: bashcompinit incorrectly calculates COMP_LINE (proposed fix included)

[Peter Stephenson]

On Fri, 17 Nov 2017 20:19:25 +1100
Evan Underscore <evanunderscore@gmail.com> wrote:
> Reverting this line (i.e. using CURRENT-1 instead of CURRENT, as it was
> before this commit) fixes the bug for me, causing zsh to report 18 and 19
> for the two test lines respectively, matching Bash.

So I think so suggesting the following patch gives the correct behaviour
under all circumstances you aware of?

pws


diff --git a/Completion/bashcompinit b/Completion/bashcompinit
index 902fa88..02290a1 100644
--- a/Completion/bashcompinit
+++ b/Completion/bashcompinit
@@ -8,7 +8,7 @@ _bash_complete() {
   local -x COMP_LINE="$words"
   local -A savejobstates savejobtexts
 
-  (( COMP_POINT = 1 + ${#${(j. .)words[1,CURRENT]}} + $#QIPREFIX + $#IPREFIX + $#PREFIX ))
+  (( COMP_POINT = 1 + ${#${(j. .)words[1,CURRENT-1]}} + $#QIPREFIX + $#IPREFIX + $#PREFIX ))
   (( COMP_CWORD = CURRENT - 1))
   COMP_WORDS=( $words )
   BASH_VERSINFO=( 2 05b 0 1 release )

Re: bashcompinit incorrectly calculates COMP_LINE (proposed fix included)

[Evan Underscore]

[-- Attachment #1: Type: text/plain, Size: 1378 bytes --]

Yes, that patch works for me for the following test cases:

./demo.py new test[tab] (COMP_POINT=18)
./demo.py new test [tab] (COMP_POINT=19)
./demo.py new 你好嘚瑟[tab] (COMP_POINT=18)
./demo.py new 你好嘚瑟 [tab] (COMP_POINT=19)

Thanks.

On Mon, Nov 20, 2017 at 6:49 AM, Peter Stephenson <
p.w.stephenson@ntlworld.com> wrote:

> On Fri, 17 Nov 2017 20:19:25 +1100
> Evan Underscore <evanunderscore@gmail.com> wrote:
> > Reverting this line (i.e. using CURRENT-1 instead of CURRENT, as it was
> > before this commit) fixes the bug for me, causing zsh to report 18 and 19
> > for the two test lines respectively, matching Bash.
>
> So I think so suggesting the following patch gives the correct behaviour
> under all circumstances you aware of?
>
> pws
>
>
> diff --git a/Completion/bashcompinit b/Completion/bashcompinit
> index 902fa88..02290a1 100644
> --- a/Completion/bashcompinit
> +++ b/Completion/bashcompinit
> @@ -8,7 +8,7 @@ _bash_complete() {
>    local -x COMP_LINE="$words"
>    local -A savejobstates savejobtexts
>
> -  (( COMP_POINT = 1 + ${#${(j. .)words[1,CURRENT]}} + $#QIPREFIX +
> $#IPREFIX + $#PREFIX ))
> +  (( COMP_POINT = 1 + ${#${(j. .)words[1,CURRENT-1]}} + $#QIPREFIX +
> $#IPREFIX + $#PREFIX ))
>    (( COMP_CWORD = CURRENT - 1))
>    COMP_WORDS=( $words )
>    BASH_VERSINFO=( 2 05b 0 1 release )
>