Merge from origin/emacs-26

[emacs.git] / doc / misc / flymake.texi

\input texinfo   @c -*-texinfo; coding: utf-8 -*-
@comment %**start of header
@setfilename ../../info/flymake.info
@set VERSION 1.0
@set UPDATED June 2018
@settitle GNU Flymake @value{VERSION}
@include docstyle.texi
@syncodeindex pg cp
@syncodeindex vr cp
@syncodeindex fn cp
@comment %**end of header

@copying
This manual is for GNU Flymake (version @value{VERSION}, @value{UPDATED}),
which is a universal on-the-fly syntax checker for GNU Emacs.

Copyright @copyright{} 2004--2018 Free Software Foundation, Inc.

@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, with the Front-Cover Texts being ``A GNU Manual,''
and with the Back-Cover Texts as in (a) below.  A copy of the license
is included in the section entitled ``GNU Free Documentation License''.

(a) The FSF's Back-Cover Text is: ``You have the freedom to copy and
modify this GNU manual.''
@end quotation
@end copying

@dircategory Emacs misc features
@direntry
* Flymake: (flymake).           A universal on-the-fly syntax checker.
@end direntry

@titlepage
@title GNU Flymake
@subtitle for version @value{VERSION}, @value{UPDATED}
@author João Távora and Pavel Kobiakov(@email{pk_at_work@@yahoo.com}).
@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@contents

@ifnottex
@node Top
@top GNU Flymake
@insertcopying
@end ifnottex

@menu
* Using Flymake::
* Extending Flymake::
* The legacy Proc backend::
* GNU Free Documentation License::
* Index::
@end menu

@node Using Flymake
@chapter Using Flymake
@cindex overview of flymake
@cindex using flymake

Flymake is a universal on-the-fly buffer checker implemented as an
Emacs minor mode.  To use Flymake, you must first activate
@code{flymake-mode} by using the command @kbd{flymake-mode}.

When enabled, Flymake collects information about problems in the
buffer, called @dfn{diagnostics}, from one or more different sources,
or @dfn{backends}, and then visually annotates the buffer by
highlighting problematic buffer regions with a special face.

It also displays an overall buffer status in the mode line containing
totals for different types of diagnostics.

Syntax check is done ``on-the-fly''.  It is started whenever

@itemize @bullet
@item
@code{flymake-mode} is started, unless
@code{flymake-start-on-flymake-mode} is nil;

@item
the buffer is saved, unless @code{flymake-start-on-save-buffer} is
nil;

@item
a newline character is added to the buffer, unless
@code{flymake-start-syntax-check-on-newline} is nil;

@item
some changes were made to the buffer more than @code{0.5} seconds ago
(the delay is configurable in @code{flymake-no-changes-timeout}).
@end itemize

Syntax check can also be started manually by typing the @kbd{M-x
flymake-start @key{RET}} command.

@code{flymake-goto-next-error} and @code{flymake-goto-prev-error} are
commands that allow easy navigation to the next/previous erroneous
line, respectively.  If might be a good idea to map them to @kbd{M-n}
and @kbd{M-p} in @code{flymake-mode}, by adding to your init file:

@lisp
(define-key flymake-mode-map (kbd "M-n") 'flymake-goto-next-error)
(define-key flymake-mode-map (kbd "M-p") 'flymake-goto-prev-error)
@end lisp

Flymake is a universal syntax checker in the sense that it's easily
extended to support new backends (@pxref{Extending Flymake}).

Historically, Flymake used to accept diagnostics from a single
backend, albeit a reasonably flexible one.

This backend isn't (yet) obsolete, and so is still available as a
fallback and active by default (@pxref{The legacy Proc backend}).  It works by
selecting a syntax check tool from a preconfigured list (compiler for
C@t{++} files, @command{perl} for Perl files, etc.), and executing it in the
background, passing it a temporary file which is a copy of the current
buffer, and parsing the output for known error/warning message
patterns.

@menu
* Syntax check statuses::
* Backend exceptions::
* Customizable variables::
@end menu

@node Syntax check statuses
@section Syntax check statuses
@cindex Syntax check statuses

When enabled, Flymake displays its status in the mode line, which
provides a visual summary of diagnostic collection.  It may also hint
at certain exceptional situations encountered when communicating with
backends.

The following statuses are defined:

@multitable @columnfractions 0.25 0.75
@item [@var{nerrors} @var{nwarnings} ...]
@tab Normal operation. @var{nerrors} and @var{nwarnings} are, respectively,
the total number of errors and warnings found during the last buffer
check, for all backends. They may be followed by other totals for
other types of diagnostics (@pxref{Flymake error types}).

@item @code{Wait}
@tab  Some Flymake backends haven't reported since the last time they
where questioned.  It is reasonable to assume that this is a temporary
delay and Flymake will resume normal operation soon.

@item @code{!}
@tab All the configured Flymake backends have disabled themselves: Flymake
cannot annotate the buffer and action from the user is needed to
investigate and remedy the situation (@pxref{Backend exceptions}).

@item @code{?}
@tab There are no applicable Flymake backends for this buffer, thus Flymake
cannot annotate it.  To fix this, a user may look to extending Flymake
and add a new backend (@pxref{Extending Flymake}).

@end multitable

@node Backend exceptions
@section Backend exceptions
@cindex backend exceptions

@cindex disabled backends
@cindex backends, disabled
Some backends may take longer than others to respond or complete, and
some may decide to @emph{disable} themselves if they are not suitable
for the current buffer or encounter some unavoidable problem.  A
disabled backend is not tried again for future checks of the current
buffer.

@findex flymake-reporting-backends
@findex flymake-running-backends
@findex flymake-disabled-backends
The commands @code{flymake-reporting-backends},
@code{flymake-running-backends} and @code{flymake-disabled-backends}
show the backends currently used and those which are disabled.

@cindex reset disabled backends
Toggling @code{flymake-mode} off and on again, or invoking
@code{flymake-start} with a prefix argument is one way to reset the
disabled backend list, so that they will be tried again in the next check.

@cindex logging
@cindex flymake logging
Flymake also uses a simple logging facility for indicating important
points in the control flow.  The logging facility sends logging
messages to the @file{*Flymake log*} buffer.  The information logged
can be used for resolving various problems related to Flymake.  For
convenience, a shortcut to this buffer can be found in Flymake's menu,
accessible from the top menu bar or just left of the status indicator.

@vindex warning-minimum-log-level
@vindex warning-minimum-level
Logging output is controlled by the Emacs @code{warning-minimum-log-level}
and @code{warning-minimum-level} variables.

@node Customizable variables
@section Customizable variables
@cindex customizable variables
@cindex variables for customizing flymake

This section summarizes customization variables used for the
configuration of the Flymake user interface.

@vtable @code
@item flymake-no-changes-timeout
If any changes are made to the buffer, syntax check is automatically
started after this many seconds, unless the user makes another change,
which resets the timer.

@item flymake-start-syntax-check-on-newline
A boolean flag indicating whether to start syntax check immediately
after a newline character is inserted into the buffer.

@item flymake-start-on-flymake-mode
A boolean flag indicating whether to start syntax check immediately
after enabling @code{flymake-mode}.

@item flymake-start-on-save-buffer
A boolean flag indicating whether to start syntax check after saving
the buffer.

@item flymake-error
A custom face for highlighting regions for which an error has been
reported.

@item flymake-warning
A custom face for highlighting regions for which a warning has been
reported.

@item flymake-note
A custom face for highlighting regions for which a note has been
reported.

@item flymake-error-bitmap
A bitmap used in the fringe to mark lines for which an error has
been reported.

@item flymake-warning-bitmap
A bitmap used in the fringe to mark lines for which a warning has
been reported.

@item flymake-fringe-indicator-position
Which fringe (if any) should show the warning/error bitmaps.

@item flymake-wrap-around
If non-nil, moving to errors with @code{flymake-goto-next-error} and
@code{flymake-goto-prev-error} wraps around buffer boundaries.
@end vtable

@node Extending Flymake
@chapter Extending Flymake
@cindex extending flymake

Flymake can primarily be extended in one of two ways:

@enumerate
@item
By changing the look and feel of the annotations produced by the
different backends.

@item
By adding a new buffer-checking backend.
@end enumerate

The following sections discuss each approach in detail.

@menu
* Flymake error types::
* Backend functions::
@end menu

@node Flymake error types
@section Customizing Flymake error types
@cindex customizing error types
@cindex error types, customization

To customize the appearance of error types, set properties on the
symbols associated with each diagnostic type.  The standard diagnostic
symbols are @code{:error}, @code{:warning} and @code{:note} (though
the backend may define more, @pxref{Backend functions}).

The following properties can be set:

@itemize

@item
@cindex bitmap of diagnostic
@code{flymake-bitmap}, an image displayed in the fringe according to
@code{flymake-fringe-indicator-position}.  The value actually follows
the syntax of @code{flymake-error-bitmap} (@pxref{Customizable
variables}).  It is overridden by any @code{before-string} overlay
property.

@item
@code{flymake-overlay-control}, an alist ((@var{OVPROP} . @var{VALUE})
@var{...}) of further properties used to affect the appearance of
Flymake annotations.  With the exception of @code{category} and
@code{evaporate}, these properties are applied directly to the created
overlay (@pxref{Overlay Properties,,, elisp, The Emacs Lisp Reference
Manual}).

As an example, here's how to make diagnostics of the type @code{:note}
stand out more prominently:

@example
(push '(face . highlight) (get :note 'flymake-overlay-control))
@end example

If you push another alist entry in front, it overrides the previous
one.  So this effectively removes the face from @code{:note}
diagnostics:

@example
(push '(face . nil) (get :note 'flymake-overlay-control))
@end example

To restore the original look for @code{:note} types, empty or remove
its @code{flymake-overlay-control} property:

@example
(put :note 'flymake-overlay-control '())
@end example

@item
@cindex severity of diagnostic
@code{flymake-severity} is a non-negative integer specifying the
diagnostic's severity.  The higher the value, the more serious is the
error.  If the overlay property @code{priority} is not specified in
@code{flymake-overlay-control}, @code{flymake-severity} is used to set
it and help sort overlapping overlays.

@item
@vindex flymake-category
@code{flymake-category} is a symbol whose property list is considered
the default for missing values of any other properties.
@end itemize

@cindex predefined diagnostic types
@vindex flymake-error
@vindex flymake-warning
@vindex flymake-note
Three default diagnostic types are predefined: @code{:error},
@code{:warning}, and @code{:note}.  By default, each one of them has a
@code{flymake-category} property whose value is, respectively, the
category symbol @code{flymake-error}, @code{flymake-warning} and
@code{flymake-note}.

These category symbols' plist is where the values of customizable
variables and faces (such as @code{flymake-error-bitmap}) are found.
Thus, if you change their plists, Flymake may stop honoring these user
customizations.

The @code{flymake-category} special property is especially useful for
backends which create diagnostics objects with non-default types that
differ from an existing type by only a few properties (@pxref{Flymake
utility functions}).

As an example, consider configuring a new diagnostic type
@code{:low-priority-note} that behaves much like @code{:note}, but
without an overlay face.

@example
(put :low-priority-note 'flymake-overlay-control '((face . nil)))
(put :low-priority-note 'flymake-category 'flymake-note)
@end example

@vindex flymake-diagnostics
@vindex flymake-diagnostic-backend
@vindex flymake-diagnostic-buffer
@vindex flymake-diagnostic-text
@vindex flymake-diagnostic-beg
@vindex flymake-diagnostic-end
As you might have guessed, Flymake's annotations are implemented as
overlays (@pxref{Overlays,,, elisp, The Emacs Lisp Reference Manual}).
Along with the properties that you specify for the specific type of
diagnostic, Flymake adds the property @code{flymake-diagnostic} to
these overlays, and sets it to the object that the backend created
with @code{flymake-make-diagnostic}.

Since overlays also support arbitrary keymaps, you can use this along
with the functions @code{flymake-diagnostics} and
@code{flymake-diagnostic-text} (@pxref{Flymake utility functions}) to
create interactive annotations, such as in the following example of
binding a @code{mouse-3} event (middle mouse button click) to an
Internet search for the text of a @code{:warning} or @code{:error}.

@example
(defun my-search-for-message (event)
  (interactive "e")
  (let* ((diags (flymake-diagnostics (posn-point (event-start event))))
         (topmost-diag (car diags)))
    (eww-browse-url
       (concat
        "https://duckduckgo.com/?q="
        (replace-regexp-in-string
          " " "+" (flymake-diagnostic-text topmost-diag)))
       t)))

(dolist (type '(:warning :error))
  (push '(mouse-face . highlight) (get type 'flymake-overlay-control))
  (push `(keymap . ,(let ((map (make-sparse-keymap)))
                      (define-key map [mouse-2]
                        'my-search-for-message)
                      map))
        (get type 'flymake-overlay-control)))
@end example

@node Backend functions
@section Backend functions
@cindex backend functions

@vindex flymake-diagnostic-functions
Flymake backends are Lisp functions placed in the special hook
@code{flymake-diagnostic-functions}.

A backend's responsibility is to diagnose the contents of a buffer for
problems, registering the problem's positions, type, and summary
description.  This information is collected in the form of diagnostic
objects created by the function @code{flymake-make-diagnostic}
(@pxref{Flymake utility functions}), and
then handed over to Flymake, which proceeds to annotate the
buffer.

A request for a buffer check, and the subsequent delivery of
diagnostics, are two key events of the interaction between Flymake
and backend.  Each such event corresponds to a well-defined function
calling convention: one for calls made by Flymake into the backend via
the backend function, the other in the reverse direction via a
callback.  To be usable, backends must adhere to both.

The first argument passed to a backend function is always
@var{report-fn}, a callback function detailed below.  Beyond it,
functions must be prepared to accept (and possibly ignore) an
arbitrary number of keyword-value pairs of the form
@w{@code{(@var{:key} @var{value} @var{:key2} @var{value2}...)}}.

Currently, Flymake may pass the following keywords and values to the
backend function:

@itemize

@item @code{:recent-changes}
The value is a list recent changes since the last time the backend
function was called for the buffer.  If the list is empty, this
indicates that no changes have been recorded.  If it is the first time
that this backend function is called for this activation of
@code{flymake-mode}, then this argument isn't provided at all
(i.e. it's not merely nil).

Each element is in the form (@var{beg} @var{end} @var{text}) where
@var{beg} and @var{end} are buffer positions, and @var{text} is a
string containing the text contained between those positions (if any),
after the change was performed.

@item @code{:changes-start} and @code{:changes-end}
The value is, repectively, the minimum and maximum buffer positions
touched by the recent changes.  These are provided for convenience and
only if @code{:recent-changes} is also provided.

@end itemize

Whenever Flymake or the user decide to re-check the buffer, backend
functions are called as detailed above, and are expected to initiate
this check, but aren't in any way required to complete it before
exiting: if the computation involved is expensive, as
is often the case with large buffers, that slower task should be
scheduled for the future using asynchronous sub-processes
(@pxref{Asynchronous Processes,,, elisp, The Emacs Lisp reference
manual}) or other asynchronous mechanisms.

In any case, backend functions are expected to return quickly or
signal an error, in which case the backend is disabled
(@pxref{Backend exceptions}).

If the function returns, Flymake considers the backend to be
@dfn{running}.  If it has not done so already, the backend is expected
to call the function @var{report-fn} passed to it, at which point
Flymake considers the backend to be @dfn{reporting}.  Backends call
@var{report-fn} by passing it a single argument @var{report-action}
followed by an optional list of keyword-value pairs of the form
@w{@code{(@var{:report-key} @var{value} @var{:report-key2} @var{value2}...)}}.

Currently accepted values for @var{report-action} are:

@itemize
@item
A (possibly empty) list of diagnostic objects created by
@code{flymake-make-diagnostic}, causing Flymake to annotate the
buffer with this information.

A backend may call @var{report-fn} repeatedly in this manner, but only
until Flymake considers that the most recently requested buffer check
is now obsolete, because, say, buffer contents have changed in the
meantime.  The backend is only given notice of this via a renewed call
to the backend function.  Thus, to prevent making obsolete reports and
wasting resources, backend functions should first cancel any ongoing
processing from previous calls.

@item
The symbol @code{:panic}, signaling that the backend has encountered
an exceptional situation and should be disabled.
@end itemize

Currently accepted @var{report-key} arguments are:

@itemize
@item
@code{:explanation}, whose value should give user-readable
details of the situation encountered, if any.

@item
@code{:force}, whose value should be a boolean suggesting
that Flymake consider the report even if it was somehow
unexpected.

@item
@code{:region}, a cons (@var{beg} . @var{end}) of buffer positions
indicating that the report applies to that region and that previous
reports targeting other parts of the buffer remain valid.
@end itemize

@menu
* Flymake utility functions::
* An annotated example backend::
@end menu

@node Flymake utility functions
@subsection Flymake utility functions
@cindex utility functions

@cindex create diagnostic object
Before delivering them to Flymake, backends create diagnostic objects
by calling the function @code{flymake-make-diagnostic}.

@deffn Function flymake-make-diagnostic buffer beg end type text
Make a Flymake diagnostic for @var{buffer}'s region from @var{beg} to
@var{end}.  @var{type} is a diagnostic symbol (@pxref{Flymake error
types}), and @var{text} is a description of the problem detected in
this region.
@end deffn

@cindex access diagnostic object
These objects' properties can be accessed with the functions
@code{flymake-diagnostic-backend}, @code{flymake-diagnostic-buffer},
@code{flymake-diagnostic-text}, @code{flymake-diagnostic-beg},
@code{flymake-diagnostic-end} and @code{flymake-diagnostic-type}.

Additionally, the function @code{flymake-diagnostics} will collect
such objects in the region you specify.

@cindex collect diagnostic objects
@deffn Function flymake-diagnostics beg end
Get a list of Flymake diagnostics in the region determined by
@var{beg} and @var{end}.  If neither @var{beg} or @var{end} is
supplied, use the whole buffer, otherwise if @var{beg} is
non-@code{nil} and @var{end} is @code{nil}, consider only diagnostics
at @var{beg}.
@end deffn

@cindex buffer position from line and column number
It is often the case with external syntax tools that a diagnostic's
position is reported in terms of a line number, and sometimes a column
number.  To convert this information into a buffer position, backends
can use the following function:

@deffn Function flymake-diag-region buffer line &optional col
Compute @var{buffer}'s region (@var{beg} .  @var{end}) corresponding
to @var{line} and @var{col}.  If @var{col} is @code{nil}, return a
region just for @var{line}.  Return @code{nil} if the region is
invalid.
@end deffn

@cindex add a log message
For troubleshooting purposes, backends may record arbitrary
exceptional or erroneous situations into the Flymake log
buffer (@pxref{Backend exceptions}):

@deffn Macro flymake-log level msg &optional args
Log, at level @var{level}, the message @var{msg} formatted with
@var{args}.  @var{level} is passed to @code{display-warning}
(@pxref{Warning Basics,,, elisp, The Emacs Lisp reference Manual}), which is
used to display the warning in Flymake's log buffer.
@end deffn

@node An annotated example backend
@subsection An annotated example backend
@cindex example of backend
@cindex backend, annotated example

This section presents an annotated example of a complete working
Flymake backend.  The example illustrates the process of writing a
backend as outlined above.

The backend in question is used for checking Ruby source files.  It
uses asynchronous sub-processes (@pxref{Asynchronous Processes,,, elisp,
The Emacs Lisp Reference Manual}), a common technique for performing
parallel processing in Emacs.

The following code needs lexical binding (@pxref{Using Lexical
Binding,,, elisp, The Emacs Lisp Reference Manual}) to be active.

@example
;;; ruby-flymake.el --- A ruby Flymake backend  -*- lexical-binding: t; -*-
(defvar-local ruby--flymake-proc nil)

(defun ruby-flymake (report-fn &rest _args)
  ;; Not having a ruby interpreter is a serious problem which should cause
  ;; the backend to disable itself, so an @code{error} is signaled.
  ;;
  (unless (executable-find
           "ruby") (error "Cannot find a suitable ruby"))
  ;; If a live process launched in an earlier check was found, that
  ;; process is killed.  When that process's sentinel eventually runs,
  ;; it will notice its obsoletion, since it have since reset
  ;; `ruby-flymake-proc' to a different value
  ;;
  (when (process-live-p ruby--flymake-proc)
    (kill-process ruby--flymake-proc))

  ;; Save the current buffer, the narrowing restriction, remove any
  ;; narrowing restriction.
  ;;
  (let ((source (current-buffer)))
    (save-restriction
      (widen)
      ;; Reset the `ruby--flymake-proc' process to a new process
      ;; calling the ruby tool.
      ;;
      (setq
       ruby--flymake-proc
       (make-process
        :name "ruby-flymake" :noquery t :connection-type 'pipe
        ;; Make output go to a temporary buffer.
        ;;
        :buffer (generate-new-buffer " *ruby-flymake*")
        :command '("ruby" "-w" "-c")
        :sentinel
        (lambda (proc _event)
          ;; Check that the process has indeed exited, as it might
          ;; be simply suspended.
          ;;
          (when (eq 'exit (process-status proc))
            (unwind-protect
                ;; Only proceed if `proc' is the same as
                ;; `ruby--flymake-proc', which indicates that
                ;; `proc' is not an obsolete process.
                ;;
                (if (with-current-buffer source (eq proc ruby--flymake-proc))
                    (with-current-buffer (process-buffer proc)
                      (goto-char (point-min))
                      ;; Parse the output buffer for diagnostic's
                      ;; messages and locations, collect them in a list
                      ;; of objects, and call `report-fn'.
                      ;;
                      (cl-loop
                       while (search-forward-regexp
                              "^\\(?:.*.rb\\|-\\):\\([0-9]+\\): \\(.*\\)$"
                              nil t)
                       for msg = (match-string 2)
                       for (beg . end) = (flymake-diag-region
                                          source
                                          (string-to-number (match-string 1)))
                       for type = (if (string-match "^warning" msg)
                                      :warning
                                    :error)
                       collect (flymake-make-diagnostic source
                                                        beg
                                                        end
                                                        type
                                                        msg)
                       into diags
                       finally (funcall report-fn diags)))
                  (flymake-log :warning "Canceling obsolete check %s"
                               proc))
              ;; Cleanup the temporary buffer used to hold the
              ;; check's output.
              ;;
              (kill-buffer (process-buffer proc)))))))
      ;; Send the buffer contents to the process's stdin, followed by
      ;; an EOF.
      ;;
      (process-send-region ruby--flymake-proc (point-min) (point-max))
      (process-send-eof ruby--flymake-proc))))

(defun ruby-setup-flymake-backend ()
  (add-hook 'flymake-diagnostic-functions 'ruby-flymake nil t))

(add-hook 'ruby-mode-hook 'ruby-setup-flymake-backend)
@end example

@node The legacy Proc backend
@chapter The legacy ``Proc'' backend
@cindex legacy proc backend

@menu
* Proc customization variables::
* Adding support for a new syntax check tool::
* Implementation overview::
* Making a temporary copy::
* Locating a master file::
* Getting the include directories::
* Locating the buildfile::
* Starting the syntax check process::
* Parsing the output::
* Interaction with other modes::
@end menu

@findex flymake-proc-legacy-backend
The backend @code{flymake-proc-legacy-backend} was originally designed
to be extended for supporting new syntax check tools and error message
patterns.  It is also controlled by its own set of customization variables

@node Proc customization variables
@section Customization variables for the Proc backend
@cindex proc backend customization variables

@vtable @code
@item flymake-proc-allowed-file-name-masks
A list of @code{(filename-regexp, init-function, cleanup-function
getfname-function)} for configuring syntax check tools.  @xref{Adding
support for a new syntax check tool}.

@item flymake-proc-master-file-dirs
A list of directories for searching a master file.  @xref{Locating a
master file}.

@item flymake-proc-get-project-include-dirs-function
A function used for obtaining a list of project include dirs (C/C++
specific).  @xref{Getting the include directories}.

@item flymake-proc-master-file-count-limit
@itemx flymake-proc-check-file-limit
Used when looking for a master file.  @xref{Locating a master file}.

@item flymake-proc-err-line-patterns
Patterns for error/warning messages in the form @code{(regexp file-idx
line-idx col-idx err-text-idx)}.  @xref{Parsing the output}.

@item flymake-proc-diagnostic-type-pred
A function to classify a diagnostic text as particular type of error.
Should be a function taking an error text and returning a diagnostic
symbol (@pxref{Flymake error types}).  If non-nil is returned but
there is no such symbol in that table, a warning is assumed.  If nil
is returned, an error is assumed.  Can also be a regular expression
that should match only warnings.  This variable replaces the old
@code{flymake-warning-re} and @code{flymake-warning-predicate}.

@item flymake-proc-compilation-prevents-syntax-check
A flag indicating whether compilation and syntax check of the same
file cannot be run simultaneously.  @xref{Interaction with other modes}.
@end vtable

@node Adding support for a new syntax check tool
@section Adding support for a new syntax check tool
@cindex adding support for a new syntax check tool

@menu
* Example---Configuring a tool called directly::
* Example---Configuring a tool called via make::
@end menu

Syntax check tools are configured using the
@code{flymake-proc-allowed-file-name-masks} list.  Each item of this list
has the following format:

@lisp
(filename-regexp, init-function, cleanup-function, getfname-function)
@end lisp

@table @code
@item filename-regexp
This field is used as a key for locating init/cleanup/getfname
functions for the buffer.  Items in
@code{flymake-proc-allowed-file-name-masks} are searched sequentially.
The first item with @code{filename-regexp} matching buffer filename is
selected.  If no match is found, @code{flymake-mode} is switched off.

@item init-function
@code{init-function} is required to initialize the syntax check,
usually by creating a temporary copy of the buffer contents.  The
function must return @code{(list cmd-name arg-list)}.  If
@code{init-function} returns null, syntax check is aborted, but
@code{flymake-mode} is not switched off.

@item cleanup-function
@code{cleanup-function} is called after the syntax check process is
complete and should take care of proper deinitialization, which is
usually deleting a temporary copy created by the @code{init-function}.

@item getfname-function
This function is used for translating filenames reported by the syntax
check tool into ``real'' filenames.  Filenames reported by the tool
will be different from the real ones, as actually the tool works with
the temporary copy.  In most cases, the default implementation
provided by Flymake, @code{flymake-proc-get-real-file-name}, can be
used as @code{getfname-function}.
@end table

To add support for a new syntax check tool, write the corresponding
@code{init-function} and, optionally, @code{cleanup-function} and
@code{getfname-function}.  If the format of error messages reported by
the new tool is not yet supported by Flymake, add a new entry to
the @code{flymake-proc-err-line-patterns} list.

The following sections contain some examples of configuring Flymake
support for various syntax check tools.

@node Example---Configuring a tool called directly
@subsection Example---Configuring a tool called directly
@cindex adding support for perl

In this example, we will add support for @command{perl} as a syntax check
tool.  @command{perl} supports the @option{-c} option which does syntax
checking.

First, we write the @code{init-function}:

@lisp
(defun flymake-proc-perl-init ()
  (let* ((temp-file (flymake-proc-init-create-temp-buffer-copy
                     'flymake-proc-create-temp-inplace))
         (local-file (file-relative-name
                      temp-file
                      (file-name-directory buffer-file-name))))
    (list "perl" (list "-wc " local-file))))
@end lisp

@code{flymake-proc-perl-init} creates a temporary copy of the buffer
contents with the help of
@code{flymake-proc-init-create-temp-buffer-copy}, and builds an appropriate
command line.

Next, we add a new entry to the
@code{flymake-proc-allowed-file-name-masks}:

@lisp
(setq flymake-proc-allowed-file-name-masks
      (cons '(".+\\.pl$"
              flymake-proc-perl-init
              flymake-proc-simple-cleanup
              flymake-proc-get-real-file-name)
            flymake-proc-allowed-file-name-masks))
@end lisp

Note that we use standard @code{cleanup-function} and
@code{getfname-function}.

Finally, we add an entry to @code{flymake-proc-err-line-patterns}:

@lisp
(setq flymake-proc-err-line-patterns
      (cons '("\\(.*\\) at \\([^ \n]+\\) line \\([0-9]+\\)[,.\n]"
              2 3 nil 1)
            flymake-proc-err-line-patterns))
@end lisp

@node Example---Configuring a tool called via make
@subsection Example---Configuring a tool called via make
@cindex adding support for C (gcc+make)

In this example we will add support for C files syntax checked by
@command{gcc} called via @command{make}.

We're not required to write any new functions, as Flymake already has
functions for @command{make}.  We just add a new entry to the
@code{flymake-proc-allowed-file-name-masks}:

@lisp
(setq flymake-proc-allowed-file-name-masks
      (cons '(".+\\.c$"
              flymake-proc-simple-make-init
              flymake-proc-simple-cleanup
              flymake-proc-get-real-file-name)
            flymake-proc-allowed-file-name-masks))
@end lisp

@code{flymake-proc-simple-make-init} builds the following @command{make}
command line:

@lisp
(list "make"
      (list "-s" "-C"
            base-dir
            (concat "CHK_SOURCES=" source)
            "SYNTAX_CHECK_MODE=1"
            "check-syntax"))
@end lisp

@code{base-dir} is a directory containing the @file{Makefile}, see
@ref{Locating the buildfile}.

Thus, @file{Makefile} must contain the @code{check-syntax} target.  In
our case this target might look like this:

@verbatim
check-syntax:
        gcc -o /dev/null -S ${CHK_SOURCES} || true
@end verbatim

@noindent
The format of error messages reported by @command{gcc} is already
supported by Flymake, so we don't have to add a new entry to
@code{flymake-err-line-patterns}.  Note that if you are using
Automake, you may want to replace @code{gcc} with the standard
Automake variable @code{COMPILE}:

@verbatim
check-syntax:
        $(COMPILE) -o /dev/null -S ${CHK_SOURCES} || true
@end verbatim

@node Implementation overview
@section Implementation overview
@cindex syntax check models
@cindex master file

@code{flymake-proc-legacy-backend} saves a copy of the buffer in a
temporary file in the buffer's directory (or in the system temporary
directory, for Java files), creates a syntax check command and
launches a process with this command.  The output is parsed using a
list of error message patterns, and error information (file name, line
number, type and text) is saved.  After the process has finished,
Flymake highlights erroneous lines in the buffer using the accumulated
error information.

Syntax check is considered possible if there's an entry in
@code{flymake-proc-allowed-file-name-masks} matching buffer's filename and
its @code{init-function} returns non-@code{nil} value.

Two syntax check modes are distinguished:

@enumerate

@item
Buffer can be syntax checked in a standalone fashion, that is, the
file (its temporary copy, in fact) can be passed over to the compiler to
do the syntax check.  Examples are C/C@t{++} sources (@file{.c},
@file{.cpp}) and Java (@file{.java}).

@item
Buffer can be syntax checked, but additional file, called master file,
is required to perform this operation.  A master file is a file that
includes the current file, so that running a syntax check tool on it
will also check syntax in the current file.  Examples are C/C@t{++}
headers (@file{.h}, @file{.hpp}).

@end enumerate

These modes are handled inside init/cleanup/getfname functions, see
@ref{Adding support for a new syntax check tool}.

The Proc backend contains implementations of all functionality
required to support different syntax check modes described above
(making temporary copies, finding master files, etc.), as well as some
tool-specific (routines for Make, Ant, etc.)@: code.


@node Making a temporary copy
@section Making a temporary copy
@cindex temporary copy of the buffer

After the possibility of the syntax check has been determined, a
temporary copy of the current buffer is made so that the most recent
unsaved changes could be seen by the syntax check tool.  Making a copy
is quite straightforward in a standalone case (mode @code{1}), as it's
just saving buffer contents to a temporary file.

Things get trickier, however, when master file is involved, as it
requires to

@itemize @bullet
@item
locate a master file
@item
patch it to include the current file using its new (temporary) name.
@end itemize

Locating a master file is discussed in the following section.

Patching just changes all appropriate lines of the master file so that they
use the new (temporary) name of the current file.  For example, suppose current
file name is @code{file.h}, the master file is @code{file.cpp}, and
it includes current file via @code{#include "file.h"}.  Current file's copy
is saved to file @code{file_flymake.h}, so the include line must be
changed to @code{#include "file_flymake.h"}.  Finally, patched master file
is saved to @code{file_flymake_master.cpp}, and the last one is passed to
the syntax check tool.

@node Locating a master file
@section Locating a master file
@cindex locating a master file
@cindex master file, locating

Master file is located in two steps.

First, a list of possible master files is built.  A simple name
matching is used to find the files.  For a C++ header @file{file.h},
the Proc backend searches for all @file{.cpp} files in the directories
whose relative paths are stored in a customizable variable
@code{flymake-proc-master-file-dirs}, which usually contains something
like @code{("." "./src")}.  No more than
@code{flymake-proc-master-file-count-limit} entries is added to the
master file list.  The list is then sorted to move files with names
@file{file.cpp} to the top.

Next, each master file in a list is checked to contain the appropriate
include directives.  No more than @code{flymake-proc-check-file-limit} of each
file are parsed.

For @file{file.h}, the include directives to look for are
@code{#include "file.h"}, @code{#include "../file.h"}, etc.  Each
include is checked against a list of include directories
(see @ref{Getting the include directories}) to be sure it points to the
correct @file{file.h}.

First matching master file found stops the search.  The master file is then
patched and saved to disk.  In case no master file is found, syntax check is
aborted, and corresponding status (@samp{!}) is reported in the mode line.
@xref{Syntax check statuses}.

@node Getting the include directories
@section Getting the include directories
@cindex include directories (C/C++ specific)

Two sets of include directories are distinguished: system include directories
and project include directories.  The former is just the contents of the
@code{INCLUDE} environment variable.  The latter is not so easy to obtain,
and the way it can be obtained can vary greatly for different projects.
Therefore, a customizable variable
@code{flymake-proc-get-project-include-dirs-function} is used to provide the
way to implement the desired behavior.

The default implementation, @code{flymake-proc-get-project-include-dirs-imp},
uses a @command{make} call.  This requires a correct base directory, that is, a
directory containing a correct @file{Makefile}, to be determined.

As obtaining the project include directories might be a costly operation, its
return value is cached in the hash table.  The cache is cleared in the beginning
of every syntax check attempt.

@node Locating the buildfile
@section Locating the buildfile
@cindex locating the buildfile
@cindex buildfile, locating
@cindex Makefile, locating

The Proc backend can be configured to use different tools for
performing syntax checks.  For example, it can use direct compiler
call to syntax check a perl script or a call to @command{make} for a
more complicated case of a C/C@t{++} source.  The general idea is
that simple files, like Perl scripts and @acronym{HTML} pages, can be checked by
directly invoking a corresponding tool.  Files that are usually more
complex and generally used as part of larger projects, might require
non-trivial options to be passed to the syntax check tool, like
include directories for C@t{++}.  The latter files are syntax checked
using some build tool, like Make or Ant.

All Make configuration data is usually stored in a file called
@file{Makefile}.  To allow for future extensions, Flymake uses a notion of
buildfile to reference the @dfn{project configuration} file.

@findex flymake-proc-find-buildfile
Special function, @code{flymake-proc-find-buildfile} is provided for locating buildfiles.
Searching for a buildfile is done in a manner similar to that of searching
for possible master files.
@ignore
A customizable variable
@code{flymake-proc-buildfile-dirs} holds a list of relative paths to the
buildfile.  They are checked sequentially until a buildfile is found.
@end ignore
In case there's no build file, the syntax check is aborted.

Buildfile values are also cached.

@node Starting the syntax check process
@section Starting the syntax check process
@cindex syntax check process, legacy proc backend

The command line (command name and the list of arguments) for
launching a process is returned by the initialization function.  The
Proc backend then just starts an asynchronous process and configures a
process filter and sentinel, which are used for processing the output
of the syntax check tool.  When exiting Emacs, running processes will
be killed without prompting the user.

@node Parsing the output
@section Parsing the output
@cindex parsing the output, legacy proc backend

The output generated by the syntax check tool is parsed in the process
filter/sentinel using the error message patterns stored in the
@code{flymake-proc-err-line-patterns} variable.  This variable contains a
list of items of the form @w{@code{(regexp file-idx line-idx
err-text-idx)}}, used to determine whether a particular line is an
error message and extract file name, line number and error text,
respectively.  Error type (error/warning) is also guessed by matching
error text with the @samp{^[wW]arning} pattern.  Anything that was not
classified as a warning is considered an error.  Type is then used to
sort error menu items, which shows error messages first.

The Proc backend is also able to interpret error message patterns
missing err-text-idx information.  This is done by merely taking the
rest of the matched line (@code{(substring line (match-end 0))}) as
error text.  This trick allows making use of a huge collection of
error message line patterns from @file{compile.el}.  All these error
patterns are appended to the end of
@code{flymake-proc-err-line-patterns}.

The error information obtained is saved in a buffer local
variable.  The buffer for which the process output belongs is
determined from the process-id@w{}->@w{}buffer mapping updated
after every process launch/exit.

@node Interaction with other modes
@section Interaction with other modes
@cindex interaction with other modes, legacy proc backend
@cindex interaction with compile mode, legacy proc backend

The only mode the Proc backend currently knows about is
@code{compile}.

The Proc backend can be configured to not start syntax check if it
thinks the compilation is in progress, by testing the
@code{compilation-in-progress} variable.  The reason why this might be
useful is saving CPU time in case both syntax check and compilation
are very CPU intensive.  The original reason for adding this feature,
though, was working around a locking problem with MS Visual C++
compiler.  The variable in question is
@code{flymake-proc-compilation-prevents-syntax-check}.

@findex flymake-proc-compile
The Proc backend also provides an alternative command for starting
compilation, @code{flymake-proc-compile}.  It just kills all the active
syntax check processes before calling @code{compile}.

@node GNU Free Documentation License
@appendix GNU Free Documentation License
@include doclicense.texi

@node Index
@unnumbered Index

@printindex cp

@bye