"replace-match-maybe-edit: Match data clobbered by buffer modification hooks" with lsp-clangd

[svenpanne]

This is basically the same issue as #1293 and #2081, this time when using query-replace in a C++ buffer with the lsp-clangd backend enabled. To reproduce: Use (c-c++ :variables c-c++-backend 'lsp-clangd) in Spacemacs' dotspacemacs-configuration-layers and do a query-replace. (lsp-mode-20201008.1601)

Probably yet another missing save-match-data, but I've got no clue where to add this...

[yyoncho]

can you provide callstack?

[svenpanne]

can you provide callstack?

How can I produce one? The *Messages* buffer contains only a one-line message as seen in the topic of this issue.

[yyoncho]

In the Report bug template there is an instruction on how to do that: https://github.com/emacs-lsp/lsp-mode/blob/master/.github/ISSUE_TEMPLATE/bug_report.md

[svenpanne]

Debugger entered--Lisp error: (error "Match data clobbered by buffer modification hooks")
  replace-match("main_reply_queue_" nil t)
  replace-match-maybe-edit("main_reply_queue_" nil t nil (5915 5921 #<buffer main.cc>) nil)
  perform-replace(#("queue_" 0 5 (face font-lock-variable-name-face) 5 6 (rear-nonsticky t face font-lock-variable-name-face)) "main_reply_queue_" t nil nil nil nil nil nil nil nil)
  query-replace(#("queue_" 0 5 (face font-lock-variable-name-face) 5 6 (rear-nonsticky t face font-lock-variable-name-face)) "main_reply_queue_" nil nil nil nil nil)
  funcall-interactively(query-replace #("queue_" 0 5 (face font-lock-variable-name-face) 5 6 (rear-nonsticky t face font-lock-variable-name-face)) "main_reply_queue_" nil nil nil nil nil)
  call-interactively(query-replace nil nil)
  command-execute(query-replace)

[svenpanne]

Is any info still missing? I don't think so, but I can't remove the need-info label, because I'm not a project owner...

[svenpanne]

The issue is still there with the latest lsp-mode-20201027.850. Is there anything else I can do to help debugging this? A broken query-replace is quite annoying in day-to-day work. :cry: Everything else is very usable, so this is a pity...

[yyoncho]

I am unable to reproduce with the provided info. Also, I do not see lsp-mode in the error callstack. Are you able to reproduce with lsp-start-plain.el?

[svenpanne]

I am unable to reproduce with the provided info. Also, I do not see lsp-mode in the error callstack.

Neither do I, but nevertheless I am relatively sure that lsp-mode is the culprit: My current workaround is to switch to any non-lsp-mode (e.g. fundamental), do the query-replace, and switch back to c/c++-mode.

I am using Spacemacs (development branch, updating almost daily) with Emacs 27.1. The relevant snippet from my .spacemacs in the dotspacemacs-configuration-layers part:

     (c-c++ :variables
            c-c++-backend 'lsp-clangd
            c-c++-default-mode-for-headers 'c++-mode
            lsp-clients-clangd-executable "clangd-11"
            company-clang-executable "clang-11"
            clang-format-executable "clang-format-11")

Are you able to reproduce with lsp-start-plain.el?

How can I do that? I'm happy to try, but I've never heard of lsp-start-plain.el...

It really looks like another save-match-data is missing somewhere in lsp-mode, but I can't really guess where. Can you give me some hints/places where I can try to add this?

[yyoncho]

How can I do that? I'm happy to try, but I've never heard of lsp-start-plain.el...

The instructions are in the bug template.

It really looks like another save-match-data is missing somewhere in lsp-mode, but I can't really guess where. Can you give me some hints/places where I can try to add this?

I have no clue.

My current workaround is to switch to any non-lsp-mode (e.g. fundamental), do the query-replace, and switch back to c/c++-mode.

Try lsp-disconnect

The relevant snippet from my .spacemacs in the dotspacemacs-configuration-layers part:

I am using spacemacs as well. The issue might be from a different layer/setting.

[svenpanne]

I am quite sure by now that the culprit is the "info box" (don't know what it is called in lsp-mode) popping up in the upper right, showing information like the signature/comments/... of the C++ entity below the cursor:

• If this pops up during query-replace, the match data is clobbered.
• If it pops up during repeatedly pressing M-d (kill-word), the kill ring is somehow broken (it did not accumulate the consecutive kills anymore).
• If it pops up during incremental search (isearch-forward, C-s), it breaks the "search again" functionality of C-s, i.e. you have to press C-s an additional time.

All of this seems to imply that the "info box" incorrectly modifies various global state which should better be left untouched.

To see all those bugs you need a working C++ project where the "info box" actually pops up all the time when the cursor is over a C++ entity. This means that you need a compile-commands.json file, have set the project root correctly etc. This might explain why you don't see the bugs I am describing. But setting up such a project should not be a problem for someone who has written lsp-clangd.el. :wink: I am a bit clueless how to give more hints with a simpler setup, but given the amount of bugs associated with the "info box", it shouldn't be too hard to reproduce on your side. :thinking:

One final thing: I update my Spacemacs packages almost every day, so the problem should be reproducible in lsp-mode's HEAD, I guess. Currently my Spacemacs uses .emacs.d/elpa/27.1/develop/lsp-mode-20201103.2013.

[wkenyon]

Still an issue for me on lsp-mode-20210330.1510

[wkenyon]

It seems to be due to the "info box" that svenpanne mentioned. My workaround is to use these functions to disable the info box (lsp-ui-doc-mode) for all buffers before doing a rename, and then reenable when I'm done renaming.

(defun w/disable-doc ()
(interactive)
  (setq lsp-ui-doc-enable nil)
  (dolist (buffer (buffer-list))
    (with-current-buffer buffer
      (lsp-ui-doc-mode -1))))

(defun w/enable-doc ()
(interactive)
  (setq lsp-ui-doc-enable 1)
  (dolist (buffer (buffer-list))
    (with-current-buffer buffer
      (lsp-ui-doc-mode 1))))

Hope it's useful to someone. p.s. lsp-mode is amazing, don't know how I lived without it. Thanks for all your hard work on it.

[squiter]

Just to bring more evidence to the problem, I'm having this exact same problem using lsp-mode for elixir.

I tried this last snippet from @wkenyon but didn't work to me, the only workaround that was working is to change the major mode to something else, do a query-replace and then go back to the elixir-mode.

[squiter]

Still an issue to me on LSP :: lsp-mode 20220429.647, Emacs 27.2, gnu/linux

[squiter]

Hey folks, just to give you some feedback, my problem wasn't related to lsp-mode, was related to the auto-highlight-symbol package.

[ianyepan]

Any updates on this issue? I'm still seeing errors despite disabling lsp-mode (clangd) and highlight-symbol-mode (similar to auto highlight symbol).