highlight groups PmenuMatch and PmenuMatchSel some buggy about vim

[Shane-XB-Qian]

@glepnir if ok, let me know if was intended to design like that, then let it be..

[Shane-XB-Qian]

and @chrisbra what do you think?
i supposed this two hl groups fit for genenral compl, not regarding to fuzzy or something as seems you said somewhere.

[glepnir]

emm a simple reprodcue script will be useful .

[Shane-XB-Qian]

and @chrisbra what do you think?
i supposed this two hl groups fit for any compl, and not regarding to fuzzy or something, as seems you said somewhere.

[glepnir]

diff --git a/src/insexpand.c b/src/insexpand.c
index 28e4d344e..c673df927 100644
--- a/src/insexpand.c
+++ b/src/insexpand.c
@@ -1448,7 +1448,7 @@ ins_compl_show_pum(void)
     char_u *
 ins_compl_leader(void)
 {
-    return compl_leader;
+    return compl_leader != NULL ? compl_leader : compl_orig_text;
 }

 /*

could you try this patch.

[Shane-XB-Qian]

and @chrisbra what do you think?
i supposed this two hl groups fit for any compl, and not regarding to fuzzy or something, as seems you said somewhere.

[glepnir]

you publish duplicate comments.

[Shane-XB-Qian]

you publish duplicate comments.

yea~ sry, we are living in the great cn, sry for the wonderful network~ 😅

[Shane-XB-Qian]

could you try this patch.

yes, the item 2 seems worked after this patch
// item 1 remained.

[zeertzjq]

It may be necessary to store the start and the end of the matched position when using non-fuzzy completion, so that highlighting of matched text works when using 'ignorecase', but I'm not entirely sure about that. That'll also be required to highlight matched text for cmdline non-fuzzy completion, as in cmdline completion the pattern can contain wildcards.

Speaking of that, should matched text be highlighted for cmdline completion as well?

[glepnir]

okay good to know. I am still finding a way to reproduce the peoblem 1.. could you provide more information about problem 1

[Shane-XB-Qian]

i have not tried to use fuzzy of completeopt yet, but generally fuzzy should be same like the case ignored.
// it is not the caseignore, but the icase of compl items attribute.

[glepnir]

ins_compl_bs then invoke ins_compl_free it free leader like currently leader is cl both freed. so when bs a leader the ins_compl_leader get a null pointer then nothing highlighted. but the compl_orig_text keep the original value.

[glepnir]

i have not tried to use fuzzy of completeopt yet, but generally fuzzy should be same like the case ignored. // it is not the caseignore, but the icase of compl items attribute.

you mean the problem 1 is run with the completeopt no fuzzy right ?

[Shane-XB-Qian]

no, it is not regarding to fuzzy of completeopt option, or to fuzzy as common sense usually its case was not matter.
icase is a complete_item attribute.
// not sure pum in cmdline, it maybe no icase attribute.

[glepnir]

reproduced.

[glepnir]

Speaking of that, should matched text be highlighted for cmdline completion as well?

should be :)

[Shane-XB-Qian]

Speaking of that, should matched text be highlighted for cmdline completion as well?

i donot know, i am not a fan of pum in cli, i usually like its behavior as in bash cli.
// but as i know @yegappan impl it as different way like insert-compl, it should be some different?