Simple fix for index out of bounds #567
Conversation
pkukielka
force-pushed
the
pkukielka/fix-autocompletion-out-of-bounds
branch
from
51d382a
to
ee4bfb7
Compare
| @@ -158,7 +158,8 @@ class CodyAutocompleteManager { | |||
| val originalText = editor.document.getText(TextRange(offset - caretPositionInLine, offset)) | |||
|
|
|||
| val prefixStartPosition = maxOf(originalText.lastIndexOf("."), originalText.lastIndexOf(" "), 0) | |||
| if (!lookupString.isNullOrEmpty() && | |||
| if (prefixStartPosition < originalText.length && | |||
There was a problem hiding this comment.
Based on the original stack trace, this happens with empty strings
@ "".subSequence(1, 0)
java.lang.StringIndexOutOfBoundsException: begin 1, end 0, length 0
| if (prefixStartPosition < originalText.length && | |
| if (prefixStartPosition + 1 < originalText.length && |
There was a problem hiding this comment.
I was crashing on this line originalText.subSequence(prefixStartPosition + 1, originalText.length))
when originalText.length = 0 and prefixStartPosition + 1 = 1 (so prefixStartPosition = 0).
Thus I think condition I added is valid and sufficient to prevent crash.
That said it can still result in producing empty subSequence and in such case startsWith will be always try no matter what lookupString is.
I would say let's double check and merge on Monday.
There was a problem hiding this comment.
Agree the current diff fixes the case when originalText.length === 0. Still, the logic reads like the bug below
if (index < text.length) {
text.charAt(index + 1)
}My brain is just wired to not allow that kind of code.
Either way, a much cleaner fix is to use the startsWith() overload from the Kotlin stdlib that accepts a startIndex parameter, and remove .subSequence(). The original code was written in Java where that overload doesn't exist
There was a problem hiding this comment.
That would be great idea, but it Returns true if a substring of this string starting at the specified offset startIndex starts with the specified prefix.
So startIndex would affect position in lookupString while we want it to affect a prefix (originalText).
Truth to be told I see at least one more bug there and I do not fully understand what that code is supposed to do exactly, so maybe let's discuss offline.
|
LGTM 👍🏻 |
pkukielka
force-pushed
the
pkukielka/fix-autocompletion-out-of-bounds
branch
from
ee4bfb7
to
ac2e148
Compare
pkukielka
force-pushed
the
pkukielka/fix-autocompletion-out-of-bounds
branch
from
ac2e148
to
a54aeb5
Compare
Fixes #565
Test plan
Truth to be told I'm totally unable to reproduce this issue artifically.
Hopefully fix is simple enough to speak for itself.