Inclusive Motions Set to Col -1 on Zero Length Lines

[mikejmcguirk]

Describe the bug
If selecting a text object with an ending col index of zero using inclusive motions, nvim_win_set_cursor produces a Cursor position outside buffer error

To Reproduce
In an empty Rust file, type the following:

/// My end col index will be zero
// I'm here so the row is valid

The select comment.outer

Expected behavior
In update_selection, instead of

api.nvim_win_set_cursor(0, { end_row + 1, end_col - end_col_offset })

Should be something like

api.nvim_win_set_cursor(0, { end_row + 1, math.max(end_col - end_col_offset, 0) })

Output of :checkhealth nvim-treesitter

==============================================================================
nvim-treesitter: ✅

Requirements ~

• ✅ OK Neovim was compiled with tree-sitter runtime ABI version 15 (required >=13).
• ✅ OK tree-sitter 0.25.8 (/home/mjm/.cargo/bin/tree-sitter)
• ✅ OK node 22.18.0 (/home/mjm/.nvm/versions/node/v22.18.0/bin/node)
• ✅ OK tar 1.35.0 (/usr/bin/tar)
• ✅ OK curl 8.5.0 (/usr/bin/curl)
  curl 8.5.0 (x86_64-pc-linux-gnu) libcurl/8.5.0 OpenSSL/3.0.13 zlib/1.3 brotli/1.1.0 zstd/1.5.5 libidn2/2.3.7 libpsl/0.21.2 (+libidn2/2.3.7) libssh/0.10.6/openssl/zlib nghttp2/1.59.0 librtmp/2.3 OpenLDAP/2.6.7
  Release-Date: 2023-12-06, security patched: 8.5.0-2ubuntu10.6
  Protocols: dict file ftp ftps gopher gophers http https imap imaps ldap ldaps mqtt pop3 pop3s rtmp rtsp scp sftp smb smbs smtp smtps telnet tftp
  Features: alt-svc AsynchDNS brotli GSS-API HSTS HTTP2 HTTPS-proxy IDN IPv6 Kerberos Largefile libz NTLM PSL SPNEGO SSL threadsafe TLS-SRP UnixSockets zstd

OS Info ~

• machine: x86_64
• release: 6.8.0-78-generic
• sysname: Linux
• version: #78-Ubuntu SMP PREEMPT_DYNAMIC Tue Aug 12 11:34:18 UTC 2025

Install directory for parsers and queries ~

• /home/mjm/.local/share/nvim/site/
• ✅ OK is writable.
• ✅ OK is in runtimepath.

Installed languages H L F I J ~

• bash ✓ ✓ ✓ . ✓

• c ✓ ✓ ✓ ✓ ✓

• c_sharp ✓ ✓ ✓ . ✓

• css ✓ . ✓ ✓ ✓

• diff ✓ . ✓ . ✓

• ecma

• git_rebase ✓ . . . ✓

• gitattributes ✓ ✓ . . ✓

• gitcommit ✓ . . . ✓

• gitignore ✓ . . . ✓

• go ✓ ✓ ✓ ✓ ✓

• html ✓ ✓ ✓ ✓ ✓

• html_tags

• javascript ✓ ✓ ✓ ✓ ✓

• json ✓ ✓ ✓ ✓ .

• jsx

• lua ✓ ✓ ✓ ✓ ✓

• markdown ✓ . ✓ ✓ ✓

• markdown_inline ✓ . . . ✓

• perl ✓ . ✓ . ✓

• python ✓ ✓ ✓ ✓ ✓

• query ✓ ✓ ✓ ✓ ✓

• rust ✓ ✓ ✓ ✓ ✓

• sql ✓ . ✓ ✓ ✓

• tmux ✓ . . . ✓

• typescript ✓ ✓ ✓ ✓ ✓

• vim ✓ ✓ ✓ . ✓

• vimdoc ✓ . . . ✓

  Legend: H[ighlights], L[ocals], F[olds], I[ndents], In[J]ections ~

Output of nvim --version

NVIM v0.12.0-dev-1074+g58060c2340
Build type: RelWithDebInfo
LuaJIT 2.1.1753364724

Additional context

If the first line is a doc comment and there is no second line, the end index will be on row one even though it doesn't exist. Perhaps update_selection could also clamp the row to the result of nvim_buf_line_count(). But I'm not sure if the Rust parser producing a non-existent row is an upstream issue.