bug#2491: 23.0.90; Flymake too aggressive when responding to unusual error combination

[sand@blarg.net]

Background
----------

The Flymake library shipped with CVS HEAD (and probably earlier) has
the following code in `flymake-post-syntax-check':

(defun flymake-post-syntax-check (exit-status command)
  ;; [. . . elided . . .]

  (if (and (equal 0 err-count) (equal 0 warn-count))
      (if (equal 0 exit-status)
          (flymake-report-status "" "")    ; PASSED
        (if (not flymake-check-was-interrupted)
            (flymake-report-fatal-status "CFGERR"
              (format "Configuration error has occured while running %s" command))
          (flymake-report-status nil ""))) ; "STOPPED"
    (flymake-report-status (format "%d/%d" err-count warn-count) "")))

Depending on...

 ...whether the flymake check generated errors or warnings,
 ...whether the flymake check had a non-zero exit status, and
 ...whether we interrupted the flymake check,

we can generate any of four different statuses.  One of the statuses,
"CFGERR" is fatal, and turns off flymake for that buffer.  The others
are non-fatal.

Problem
-------

Flymake only tracks errors that refer to the specific file being
checked.  For example, given the following output for "foo.c",
which includes file "bar.h"

foo.c:20: warning: Type mismatch
bar.h:30: error: Invalid syntax

Flymake would "see" one warning and no errors.  The "bar.h" error is
dropped.  Now consider the degenerate case where "bar.h" is the only
source of errors, *and the check returns a non-zero exit status*:

bar.h:30: error: Invalid syntax

The error and warning counts are zero, the exit status is non-zero,
the check was not interrupted, so we end up reporting a fatal CFGERR.

Think aobut the user experience here, with the two examples shown
above.  First, Flymake checks, and reports an warning in the current
file "foo.c".  The user fixes the warning.  Then Flymake reports a
fatal error (popping up a dialog box under X) and turns itself off!

Solution
--------

That particular code branch is supposed to catch cases where the build
system itself dies with an error, which is important to detect.  But
shutting down Flymake is too excessive a response, because of errors
in included files.

The following replacement code changes the behavior to report zero
errors and zero warnings with the file itself, but it adds a ":CFGERR"
flag to indicate that there was some other problem with the check.
Flymake remains enabled for the buffer.

(defun flymake-post-syntax-check (exit-status command)
  ;; [. . . elided . . .]

  (if (and (equal 0 err-count) (equal 0 warn-count))
      (if (equal 0 exit-status)
          (flymake-report-status "" "") ; PASSED
        (if (not flymake-check-was-interrupted)
            (flymake-report-status "0/0" ":CFGERR")
          (flymake-report-status nil ""))) ; "STOPPED"
    (flymake-report-status (format "%d/%d" err-count warn-count) "")))


Derek




In GNU Emacs 23.0.90.1 (i486-pc-linux-gnu, GTK+ Version 2.12.11)
 of 2009-02-07 on elegiac, modified by Debian
 (emacs-snapshot package, version 1:20090207-1)
Windowing system distributor `The X.Org Foundation', version 11.0.10402000
configured using `configure  '--build' 'i486-linux-gnu' '--host' 'i486-linux-gnu' '--prefix=/usr' '--sharedstatedir=/var/lib' '--libexecdir=/usr/lib' '--localstatedir=/var' '--infodir=/usr/share/info' '--mandir=/usr/share/man' '--with-pop=yes' '--enable-locallisppath=/etc/emacs-snapshot:/etc/emacs:/usr/local/share/emacs/23.0.90/site-lisp:/usr/local/share/emacs/site-lisp:/usr/share/emacs/23.0.90/site-lisp:/usr/share/emacs/site-lisp' '--with-x=yes' '--with-x-toolkit=gtk' 'build_alias=i486-linux-gnu' 'host_alias=i486-linux-gnu' 'CFLAGS=-DDEBIAN -DSITELOAD_PURESIZE_EXTRA=5000 -g -O2' 'LDFLAGS=-g -Wl,--as-needed' 'CPPFLAGS=''