Introduce flycheck-last-checker to remember the last checker in flycheck-get-checker.

Add commands to jump forwards/backwards between errors.

Chaining of checkers

Enable chaining of checkers, i.e. execute checker A after checker B.

Rationale

Sometimes its useful to apply use more than one checkers to a buffer. For instance, in Haskell one'd like to first check for syntax error using ghc, and afterwards check for style errors with hslint.

Interface

Add property :next-checker. The property is a list, where each item is either a checker symbol or a cons cell (PREDICATE . SYMBOL). PREDICATE determines when to run the checker, SYMBOL refers to the checker to run. PREDICATE may be one of

• no-errors: Run only if the current checker gave no errors at all.
• warnings-only: Run if the current checker gave warnings only.

The second element is a checker symbol.

Example

(defvar flycheck-checker-haskell-ghc
   '(:command
     ...
     :next-checker
     ((only-warnings . flycheck-checker-hslint))))

Implementation

TBD.

Status

Planning.

The current flycheck calls s-numeric?, but I don't have such a function on my system.

The python checkers do no mark an error when there are syntax error on the python file. However, the underlying commands (pylint, pyflakes, etc) do report the error.

Example:

prin t "hello"

The ghc checker reports "Foo" for errors and "Warning: Foo" for warnings, but the way flycheck uses --mapcat, it thinks that both an error and a warning have occurred. Once a pattern triggers, that error output should not be used again.

When enabling flycheck in emacs-lisp-mode, it will inevitably be enabled when browsing core library code, which is compressed.

Here's what happens if I have enabled flycheck in, say, "ido.el.gz":

Debugger entered--Lisp error: (wrong-type-argument number-or-marker-p nil)
  jka-compr-write-region("" nil "/var/folders/hc/w17tmcyn47j16fs8h4rb3_6w0000gn/T/flymake27516f8S.gz" nil silent)
  apply(jka-compr-write-region ("" nil "/var/folders/hc/w17tmcyn47j16fs8h4rb3_6w0000gn/T/flymake27516f8S.gz" nil silent))
  jka-compr-handler(write-region "" nil "/var/folders/hc/w17tmcyn47j16fs8h4rb3_6w0000gn/T/flymake27516f8S.gz" nil silent)
  write-region("" nil "/var/folders/hc/w17tmcyn47j16fs8h4rb3_6w0000gn/T/flymake27516f8S.gz" nil silent nil excl)
  byte-code("<elided>" [prefix temporary-file-directory file suffix dir-flag make-temp-name zerop file-name-as-directory expand-file-name make-directory write-region "" nil silent excl] 8)
  make-temp-file("flymake" nil ".gz")
  flycheck-create-temp-system("/Applications/Emacs.app/Contents/Resources/lisp/ido.el.gz" "flymake")
  flymake-init-create-temp-buffer-copy(flycheck-create-temp-system)
  flycheck-substitute-argument(source)
  mapcar(flycheck-substitute-argument ("/Applications/Emacs.app/Contents/MacOS/Emacs" "--no-site-file" "--no-site-lisp" "--batch" "--eval" "\n(progn (setq byte-compile-dest-file-function (quote make-temp-file)) (dolist (file command-line-args-left) (byte-compile-file file)))\n" source))
  flycheck-get-substituted-command((:command ("/Applications/Emacs.app/Contents/MacOS/Emacs" "--no-site-file" "--no-site-lisp" "--batch" "--eval" "\n(progn (setq byte-compile-dest-file-function (quote make-temp-file)) (dolist (file command-line-args-left) (byte-compile-file file)))\n" source) :modes emacs-lisp-mode))
  flycheck-init-checker((:command ("/Applications/Emacs.app/Contents/MacOS/Emacs" "--no-site-file" "--no-site-lisp" "--batch" "--eval" "\n(progn (setq byte-compile-dest-file-function (quote make-temp-file)) (dolist (file command-line-args-left) (byte-compile-file file)))\n" source) :modes emacs-lisp-mode))
  byte-code("<elided>" [flycheck-checkers checker --dolist-tail-- properties nil flycheck-get-checker-properties flymake-log 3 "Trying checker %S with properties %S" flycheck-may-use-checker throw --cl-block-nil-- flycheck-init-checker] 6)
  flycheck-init()

I guess that the ".gz" extension for the temp file messes things up.

I don't load flymake in my .emacs, but now that flycheck doesn't even load flymake, flymake-errline and flymake-warnline aren't defined initially when flycheck is loaded first with a hook.

When highlighting just the error column, I'd still like to the see the error line with a very light-pink background. So, I would like there to be two error faces, but only one of them would be used in non-column highlighting mode.

Some modes have multiple checkers. For instance python-mode has three checkers using flake8,
pylint~or ~pyflakes. When doing syntax checking in python-mode, the checkers are tried in this order > and the first whose executable is found is used.

I want to use flake8 with pylint when editing a python file. According to your documentation, perhap it's impossible to do this.

I think it's a really useful features if you can implement it.

Add checkers for Haskell.

Check Haskell buffers for syntax errors first, using ghc, and afterwards check for style errors with hslint.

Depends on #31.

Just downloaded snapshot of today, the I got the following error when tried to enable flycheck-mode in python-mode.

Debugger entered--Lisp error: (void-variable let)
  (cond ((string= directory "/") (when (file-exists-p full-path) full-path)) ((file-exists-p full-path) full-path) (let ((parent-directory (file-name-directory (directory-file-name (file-name-directory full-path))))) ((flycheck-find-file-in-tree filename parent-directory))))
  (let ((full-path (expand-file-name filename directory))) (cond ((string= directory "/") (when (file-exists-p full-path) full-path)) ((file-exists-p full-path) full-path) (let ((parent-directory (file-name-directory (directory-file-name (file-name-directory full-path))))) ((flycheck-find-file-in-tree filename parent-directory)))))
  flycheck-find-file-in-tree(".jshintrc" "e:/Python/nova-2011.2/nova/")

Seems (let .... should be (t (let .... here.

But another couple of problems:

1. why .jshintrc would be searched for python-mode?
2. (after fixing the let problem) on windows, flycheck-find-file-in-tree would run into end-less recursion. I think (string= directory (file-name-directory directory)) is a better exit than (string directory "/")

Need a lot more unit tests.

Make flycheck-mode global (see define-globalized-minor-mode and :global keyword).

Allows to enable flycheck-mode in all buffers regardless of visit-file-hook.

Requirements

• Search suitable syntax checker immediately after enabling the mode.
• Disable the mode if no syntax checker is available.

Advantages

• Unified setup for all kinds of buffers, including temporary buffers without backing files.

Disadvantages

• Syntax checking is enabled for all sorts of buffers, even if it doesn't make any sense at all (.e.g help buffers, customize buffers, etc.).

I have used emacs V24.3.50.1 in mac 10.8.2 installed by :brew install emacs --cocoa --use-git-head.and configuration cloned from bbatsov/prelude master code.

When Emacs started, message buffer shows some warning like:

Checker emacs-lisp-checkdoc returned non-zero exit code 255, but no errors fromoutput: 


Args out of range: "", 1
*** #<buffer  *temp*>: checkdoc-current-buffer V 0.6.1
Checker definition probably flawed.

I am not sure where is wrong, @bbatsov deem: Seems like an upstream bug in flycheck. so i report issue here.

When I add flycheck-mode as hook,
(add-hook 'LaTeX-mode-hook 'flycheck-mode)
and creating a temp buffer (i.e. when using org-mode export temp buffer)
flycheck returns an error
flycheck-find-file-for-buffer: Wrong type argument: stringp, nil

Happens also with other checkers, like html.

In temp buffers filename is nil, so functions like
flycheck-find-file-for-buffer and flycheck-create-temp-system
don't work properly when buffer is temp buffer.

flycheck 0.5 seems to have broken a number of checkers but I can't make sense of the error messages without being able to set flymake-log-level to 3.

Use markdown for README and CONTRIBUTING, because Org formatting is really broken on Github.

I don't quite follow the reason behind delaying this information. If I hit next-error, I'm doing so because I want to know...

For example, xmllint outputs the following:

environment.plist:5: parser error : expected '>'
    <key>MAVEN_OPTS</keyx>
                        ^

In a case like this, I can't find the column number using just a regexp, but I can find it programmatically by counting the number of spaces before the caret on the third line.

Also, since there is no newline between these error message, I need some way to be able to walk through the buffer around the point of match, and then back-track to the position where the next regexp match should resume from.

Add checker for Emacs Lisp calling (checkdoc-current-buffer t) in an Emacs subprocess.

Use checker chaining #31 to call this checker after the standard Emacs Lisp.

flycheck is awesome plugin for emacs. I am writing javascript with jshint, which has so many features. Could support jshint to flycheck.

Currently each error line is matched against each error-pattern defined in the checker. When customizing checkers, it could be useful if the matching stops at the first error-pattern matched (similar to how url routers work in most web frameworks). This facilitates writing certain patterns: suppose I want flake8 warnings with a code > 800 to show as a flycheck error. I would have to write code like the one below. Note that the last error-pattern has to specifically avoid matching the previous one, otherwise the flake8 warning will be marked twice as a flycheck warning and as an flycheck error.

(flycheck-declare-checker my-python-flycheck-checker
  "my-python-flycheck-checker"
  :command '("flake8" "--max-complexity=12" source)
  :error-patterns
'(("^\\(?1:.*\\):\\(?2:[0-9]+\\): \\(?4:[[:alpha:]]\\{2\\}.*\\)$" error)
  ("^\\(?1:.*?\\):\\(?2:[0-9]+\\):\\(?:\\(?3:[0-9]*\\):?\\) \\(?4:E[0-9]+.*\\)$" error)
    ("^\\(?1:.*?\\):\\(?2:[0-9]+\\):\\(?:\\(?3:[0-9]*\\):?\\) \\(?4:W8.*\\)$" error)
    ("^\\(?1:.*?\\):\\(?2:[0-9]+\\):\\(?:\\(?3:[0-9]*\\):?\\) \\(?4:W[^8].*\\)$" warning)
    )
  :modes 'python-mode)

I installed flycheck and tried it out with some python files. I can see errors and warnings highlighted, but cant see the actual message anywhere. I'm most likely doing something wrong, but any pointers would be appreciated.

This is the behavior for 0.4, as you can see the error indicator works quite well with the fringe linum-mode added on the left. The error indicator is also put in the right place:

This is what happens for 0.5:

In addition to the error indicator added a new column to the error lines, the error lines also don't get the flymake-errline face unless turn off and then turn on flycheck-mode again.

flycheck is active in the buffers that are created during package installation. This is causing an error with the sentinel selecting a deleted buffer. I reinstalled flycheck using the repository using version
flycheck-20130127.1807

disabling flycheck fixes the issue.

As the image shows the exclamation mark icon is not shown correctly on Linux.

Allow to select a specific checker per buffer.

Introduce a buffer-local variable flycheck-checker. The value of this variable is either:

• a checker symbol
• or a checker definition

If set this checker is directly used and the standard checker discovery is skipped. If the checker cannot be used, emit a warning and do not perform a syntax check at all.

For Haskell, collapsing whitespace makes certain diagnostics almost impossible to read. I fixed this locally with:

(defalias 's-collapse-whitespace 'identity)

It would be nice if this could be configured per-checker.

Hi there!

I'm using Emacs 23.3.50.1 (Aquamacs 2.4) and installed Flycheck via MELPA and getting the following error when launching flycheck-mode:

error in process sentinal: Symbol's function definition is void: condition-case-unless-debug

Here is my config:

(require 'flycheck)
(add-hook 'ruby-mode-hook 'flycheck-mode)

Any idea what could be causing this?

Thanks,
Matt

Add support for multiline error messages, required by Emacs Lisp and probably others.

May likely require large flymake hacking, because currently flymake processes output on the fly and thus may attempt to parse incomplete messages.

Probably add :parser keyword to provide a function for parsing error messages.

Some Prelude users are getting the following error:

Warning (initialization): An error occurred while loading `/Users/akshatpradhan/.emacs.d/init.el':                                   

Symbol's function definition is void: define-fringe-bitmap                                                                           

To ensure normal operation, you should investigate and remove the                                                                    
cause of the error in your initialization file.  Start Emacs with                                                                    
the `--debug-init' option to view a complete error backtrace.  

More details here bbatsov/prelude#231 (comment)

Oddly enough I'm using the same Emacs 24.2 and flycheck works for me (meaning I have the define-fringe-bitmap function. I wonder if its a compilation flag problem or something like that.

This is what I get on the flycheck HEAD after doing M-x flymake-mode. I don't know what caused it (I suspect 25f34fa) , I can look into it if you give me some pointers.

I really like flycheck and decided to include it by default in Emacs Prelude, but some very strange problem appeared afterwards. On Emacs startup I get infinite number of the following errors:

Buffer " Compiler Input" has a running process; kill it ?

and

error in process sentinel : Selecting deleted buffer

More details here - bbatsov/prelude#248

flycheck seems to interfere with the byte-code compiler when enabled in prog-mode for some reason. Any ideas what might be causing the problem?

When updating to the latest head for the past few days, I have been getting this parsing error with the phpcs checker turned on in php-mode:

https://gist.github.com/4694396

I was using the phpcs checker fine last week.

The error goes away if I revert to this earlier commit.

3d98781 * Inhibit checkdoc in autoload source buffers [#45]

Let me know if I can provide more information about this possible bug.

For my Haskell code, I'd like to run both "hdevtools check" and "hlint". The former only, if it has errors, and the latter if it doesn't. This way I get the most checking possible.

This is how indicator displayed at here:

flycheck-error-indicator's default value is "\240"

I tried to implement the checker. This works for me in .emacs:

(flycheck-declare-checker phpcs
  "A PHP syntax checker using the PHP_CodeSniffer.

See URL `http://pear.php.net/package/PHP_CodeSniffer/'."
  :command '("phpcs" "--report=emacs" source)
  :error-patterns
  '(("\\(?1:.*\\):\\(?2:[0-9]+\\):\\(?3:[0-9]+\\): error - \\(?4:.*\\)" error)
    ("\\(?1:.*\\):\\(?2:[0-9]+\\):\\(?3:[0-9]+\\): warning - \\(?4:.*\\)" warning))
  :modes '(php-mode php+-mode)
  :next-checkers '(php))
(add-to-list 'flycheck-checkers 'phpcs)

Feel free to use it...

If I hit next-error at the end of the buffer, I'd love an option to have it automatically cycle back to the beginning.

I am using a custom syntax check tool and flycheck runs fine and finds:

... error C2143: syntax error : missing ';' before 'return'', got line-err-info

However, when the syntax check exits, flycheck seems to have lost the error:

0 error(s), 0 warning(s), 0 info in 1.15 second(s)

which results in a CFGERR?

The reason for this is that

(when (flymake-same-files real-file-name source-file-name)

fails since real-file-name has "_flymake" appended to it (I use source-inplace).

Would an Info manual be useful for Flycheck?

Opinions?

Largely replace the README with a proper Info manual.

The README will only contain a short introduction, installation and usage information, and the credits as well as the license blurp. No information on customization and extending will remain.

Advantages

• Online help within Emacs
• Cross-references from docstrings

Disadvantages

• Crude markup (not so crude actually)
• Little to no tool support
• No Github rendering
• Unfancy HTML rendering like from the 90s

Status

Writing of an Info manual has begun in the info-manual branch.

Remaining chapters

• Introduction
• Usage
• Customization
• Extending
• Further help
• Credits
• Index

Other tasks

• Write Generate top level dir file.
• Update the MELPA recipe to include the info file and dir file.
• Implement automatic Marmalade packaging.

I really hate the icon at the beginning of a warning or error line.
It's messed up my format. Is warning and error face is not enough?

Please tell me the way to remove it.

Enable parsing of errors with a function specified in the checker declaration.

Interface

(defun my-cool-checker-parse-output (properties buffer output) …)

(defvar my-cool-checker
   '(…
    :error-parse-function 'my-cool-checker-parse-output)

The function my-cool-checker-parse-output is expected to return a list of flycheck-error objects.

Rationale

Some checking tools can print errors in some structured output format like XML or JSON (.e.g coffeelint or jshint). Parsing structured output directly into error objects is more reliable than throwing regular expressions onto human-readable output.

I'm using this code right now to good effect:

(defun my-flycheck-show-error-in-window ()
  (interactive)
  (flycheck-cancel-error-display-timer)
  (when flycheck-mode
    (let ((buf (get-buffer-create "*Flycheck Info*"))
          (message (car (flycheck-overlay-messages-at (point)))))
      (with-current-buffer buf
        (delete-region (point-min) (point-max))
        (insert message))
      (display-buffer buf)
      (fit-window-to-buffer (get-buffer-window buf)))))

(defun flycheck-show-error-at-point ()
  "Show the first error message at point in minibuffer."
  (interactive)
  (flycheck-cancel-error-display-timer)
  (when flycheck-mode
    (if (flycheck-may-show-message)
        (let* ((buf (get-buffer-create "*Flycheck Info*"))
               (wind (get-buffer-window buf))
               (message (car (flycheck-overlay-messages-at (point)))))
          (if message
              (if (> (length (split-string message "\n")) 8)
                  (my-flycheck-show-error-in-window)
                (if wind (delete-window wind))
                (message "%s" message))
            (message nil)))
      ;; Try again if the minibuffer is busy at the moment
      (flycheck-show-error-at-point-soon))))

As you can see from screenshot, flycheck show me that line has a warning, but use flymake-errline face to indicate it.
Is this a bug?

In addition for that, https://github.com/illusori/emacs-flymake has a useful feature.

Error messages from the syntax checker can now be classed as info messages in addition or error and warning, and there's a new face to highlight info messages in: flymake-infoline

Do you think it's a good feature to adopt?

Using M-x flycheck-compile-buffer the checker output will not be parsed, but printed into a compilation buffer instead.

Mainly intended to debug and test new syntax checkers.

Show the message of the error at point in the minibuffer, just like flymake-cursor.

When editing a file over ssh with tramp, emacs stops responsing and wont work before kiling ssh process under emacs.

I think it tries to run syntax checker on remote which is not installed.