fix(java,rsync,ssh): complete syntactically incomplete cur #1485
Conversation
|
|
||
| local REPLY | ||
| _comp_dequote "$cur" | ||
| _comp_dequote_incomplete "$cur" |
There was a problem hiding this comment.
It's not immediately clear to me why we'd do this when completing java packages, the commit messages nor the referenced issues don't mention it either, and there's seemingly no test suite coverage for it. Is it necessary, and if yes, can we somehow clarify why and add test cases?
There was a problem hiding this comment.
With javadoc 'a[tab], cur becomes an incomplete word cur="'a". If we use _comp_dequote, nothing is completed because modules starting with 'a are not found. In dealing with a value coming from cur, we need to use _comp_dequote_incomplete in general.
- I rebased on top of the latest
main, and I added a test case intest/t/test_javadoc.py. - I described this in the commit message of the first commit.
- I mentioned it in the code documentation of
_comp_dequoteand_comp_dequote_incomplete.
akinomyoga
force-pushed
the
_comp_dequote_incomplete
branch
from
ce45cc9 to
3c28b9b
Compare
scop#1255 (comment) `cur` is the word that is currently input by the user and can be syntactically incomplete when the completion is requested. For example, it is typical to attempt a completion with an opening quotation `'` but without a closing quotation: $ javadoc 'a[tab] Or, if there are two candidates `a\ b.txt` and `a\\\ c.txt`, the first attempt of the completion would insert the common part `a\`, where the escape target of `\` is missing. To handle these cases, in dequoting values coming from `cur`, we need to deal with the cases with incomplete values. This patch adds a new utility `_comp_dequote_incomplete` and replace `_comp_dequote` with it everywhere `_comp_dequote` is used for values coming from `cur`. Co-authored-by: Yedaya Katsman <[email protected]>
akinomyoga
force-pushed
the
_comp_dequote_incomplete
branch
from
3c28b9b to
455f8c1
Compare
There was a problem hiding this comment.
Missed in the review and already merged, but I think this is a filename that will cause problems e.g. on Windows, and it would be better created dynamically.
2 participants
This is also separated from #1357. See the commit messages. This fixes #1255.