1;;; pcvs-parse.el --- the CVS output parser
2
3;; Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4;;   2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
5
6;; Author: Stefan Monnier <monnier@iro.umontreal.ca>
7;; Keywords: pcl-cvs
8
9;; This file is part of GNU Emacs.
10
11;; GNU Emacs is free software; you can redistribute it and/or modify
12;; it under the terms of the GNU General Public License as published by
13;; the Free Software Foundation; either version 2, or (at your option)
14;; any later version.
15
16;; GNU Emacs is distributed in the hope that it will be useful,
17;; but WITHOUT ANY WARRANTY; without even the implied warranty of
18;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
19;; GNU General Public License for more details.
20
21;; You should have received a copy of the GNU General Public License
22;; along with GNU Emacs; see the file COPYING.  If not, write to the
23;; Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
24;; Boston, MA 02110-1301, USA.
25
26;;; Commentary:
27
28;;; Bugs:
29
30;; - when merging a modified file, if the merge says that the file already
31;;   contained in the changes, it marks the file as `up-to-date' although
32;;   it might still contain further changes.
33;;   Example: merging a zero-change commit.
34
35;;; Code:
36
37(eval-when-compile (require 'cl))
38
39(require 'pcvs-util)
40(require 'pcvs-info)
41
42;; imported from pcvs.el
43(defvar cvs-execute-single-dir)
44
45;; parse vars
46
47(defcustom cvs-update-prog-output-skip-regexp "$"
48  "*A regexp that matches the end of the output from all cvs update programs.
49That is, output from any programs that are run by CVS (by the flag -u
50in the `modules' file - see cvs(5)) when `cvs update' is performed should
51terminate with a line that this regexp matches.  It is enough that
52some part of the line is matched.
53
54The default (a single $) fits programs without output."
55  :group 'pcl-cvs
56  :type '(regexp :value "$"))
57
58(defcustom cvs-parse-ignored-messages
59  '("Executing ssh-askpass to query the password.*$"
60    ".*Remote host denied X11 forwarding.*$")
61  "*A list of regexps matching messages that should be ignored by the parser.
62Each regexp should match a whole set of lines and should hence be terminated
63by `$'."
64  :group 'pcl-cvs
65  :type '(repeat regexp))
66
67;; a few more defvars just to shut up the compiler
68(defvar cvs-start)
69(defvar cvs-current-dir)
70(defvar cvs-current-subdir)
71(defvar dont-change-disc)
72
73;;;; The parser
74
75(defconst cvs-parse-known-commands
76  '("status" "add" "commit" "update" "remove" "checkout" "ci")
77  "List of CVS commands whose output is understood by the parser.")
78
79(defun cvs-parse-buffer (parse-spec dont-change-disc &optional subdir)
80  "Parse current buffer according to PARSE-SPEC.
81PARSE-SPEC is a function of no argument advancing the point and returning
82  either a fileinfo or t (if the matched text should be ignored) or
83  nil if it didn't match anything.
84DONT-CHANGE-DISC just indicates whether the command was changing the disc
85  or not (useful to tell the difference between `cvs-examine' and `cvs-update'
86  output.
87The path names should be interpreted as relative to SUBDIR (defaults
88  to the `default-directory').
89Return a list of collected entries, or t if an error occurred."
90  (goto-char (point-min))
91  (let ((fileinfos ())
92	(cvs-current-dir "")
93	(case-fold-search nil)
94	(cvs-current-subdir (or subdir "")))
95    (while (not (or (eobp) (eq fileinfos t)))
96      (let ((ret (cvs-parse-run-table parse-spec)))
97	(cond
98	 ;; it matched a known information message
99	 ((cvs-fileinfo-p ret) (push ret fileinfos))
100	 ;; it didn't match anything at all (impossible)
101	 ((and (consp ret) (cvs-fileinfo-p (car ret)))
102	  (setq fileinfos (append ret fileinfos)))
103	 ((null ret) (setq fileinfos t))
104	 ;; it matched something that should be ignored
105	 (t nil))))
106    (nreverse fileinfos)))
107
108
109;; All those parsing macros/functions should return a success indicator
110(defsubst cvs-parse-msg () (buffer-substring cvs-start (1- (point))))
111
112;;(defsubst COLLECT (exp) (push exp *result*))
113;;(defsubst PROG (e) t)
114;;(defmacro SEQ (&rest seqs) (cons 'and seqs))
115
116(defmacro cvs-match (re &rest matches)
117  "Try to match RE and extract submatches.
118If RE matches, advance the point until the line after the match and
119then assign the variables as specified in MATCHES (via `setq')."
120  (cons 'cvs-do-match
121	(cons re (mapcar (lambda (match)
122			   `(cons ',(first match) ,(second match)))
123			 matches))))
124
125(defun cvs-do-match (re &rest matches)
126  "Internal function for the `cvs-match' macro.
127Match RE and if successful, execute MATCHES."
128  ;; Is it a match?
129  (when (looking-at re)
130    (goto-char (match-end 0))
131    ;; Skip the newline (unless we already are at the end of the buffer).
132    (when (and (eolp) (< (point) (point-max))) (forward-char))
133    ;; assign the matches
134    (dolist (match matches t)
135      (let ((val (cdr match)))
136	(set (car match) (if (integerp val) (match-string val) val))))))
137
138(defmacro cvs-or (&rest alts)
139  "Try each one of the ALTS alternatives until one matches."
140  `(let ((-cvs-parse-point (point)))
141     ,(cons 'or
142	    (mapcar (lambda (es)
143		      `(or ,es (ignore (goto-char -cvs-parse-point))))
144		    alts))))
145(def-edebug-spec cvs-or t)
146
147;; This is how parser tables should be executed
148(defun cvs-parse-run-table (parse-spec)
149  "Run PARSE-SPEC and provide sensible default behavior."
150  (unless (bolp) (forward-line 1))	;this should never be needed
151  (let ((cvs-start (point)))
152    (cvs-or
153     (funcall parse-spec)
154
155     (dolist (re cvs-parse-ignored-messages)
156       (when (cvs-match re) (return t)))
157
158     ;; This is a parse error.  Create a message-type fileinfo.
159     (and
160      (cvs-match ".*$")
161      (cvs-create-fileinfo 'MESSAGE cvs-current-dir " "
162			   ;; (concat " Unknown msg: '"
163			   (cvs-parse-msg) ;; "'")
164			   :subtype 'ERROR)))))
165
166
167(defun cvs-parsed-fileinfo (type path &optional directory &rest keys)
168  "Create a fileinfo.
169TYPE can either be a type symbol or a cons of the form (TYPE . SUBTYPE).
170PATH is the filename.
171DIRECTORY influences the way PATH is interpreted:
172- if it's a string, it denotes the directory in which PATH (which should then be
173  a plain file name with no directory component) resides.
174- if it's nil, the PATH should not be trusted: if it has a directory
175  component, use it, else, assume it is relative to the current directory.
176- else, the PATH should be trusted to be relative to the root
177  directory (i.e. if there is no directory component, it means the file
178  is inside the main directory).
179The remaining KEYS are passed directly to `cvs-create-fileinfo'."
180  (let ((dir directory)
181	(file path))
182    ;; only trust the directory if it's a string
183    (unless (stringp directory)
184      ;; else, if the directory is true, the path should be trusted
185      (setq dir (or (file-name-directory path) (if directory "")))
186      (setq file (file-name-nondirectory path)))
187
188    (let ((type (if (consp type) (car type) type))
189	  (subtype (if (consp type) (cdr type))))
190      (when dir (setq cvs-current-dir dir))
191      (apply 'cvs-create-fileinfo type
192	     (concat cvs-current-subdir (or dir cvs-current-dir))
193	     file (cvs-parse-msg) :subtype subtype keys))))
194
195;;;; CVS Process Parser Tables:
196;;;;
197;;;; The table for status and update could actually be merged since they
198;;;; don't conflict.  But they don't overlap much either.
199
200(defun cvs-parse-table ()
201  "Table of message objects for `cvs-parse-process'."
202  (let (c file dir path base-rev subtype)
203    (cvs-or
204
205     (cvs-parse-status)
206     (cvs-parse-merge)
207     (cvs-parse-commit)
208
209     ;; this is not necessary because the fileinfo merging will remove
210     ;; such duplicate info and luckily the second info is the one we want.
211     ;; (and (cvs-match "M \\(.*\\)$" (path 1))
212     ;;      (cvs-parse-merge path))
213
214     ;; Normal file state indicator.
215     (and
216      (cvs-match "\\([MARCUPNJ?]\\) \\(.*\\)$" (c 1) (path 2))
217      ;; M: The file is modified by the user, and untouched in the repository.
218      ;; A: The file is "cvs add"ed, but not "cvs ci"ed.
219      ;; R: The file is "cvs remove"ed, but not "cvs ci"ed.
220      ;; C: Conflict
221      ;; U: The file is copied from the repository.
222      ;; P: The file was patched from the repository.
223      ;; ?: Unknown file.
224      (let ((code (aref c 0)))
225	(cvs-parsed-fileinfo
226	 (case code
227	   (?M 'MODIFIED)
228	   (?A 'ADDED)
229	   (?R 'REMOVED)
230	   (?? 'UNKNOWN)
231	   (?C
232	    (if (not dont-change-disc) 'CONFLICT
233	      ;; This is ambiguous.  We should look for conflict markers in the
234	      ;; file to decide between CONFLICT and NEED-MERGE.  With CVS-1.10
235	      ;; servers, this should not be necessary, because they return
236	      ;; a complete merge output.
237	      (with-temp-buffer
238		(insert-file-contents path)
239		(goto-char (point-min))
240		(if (re-search-forward "^<<<<<<< " nil t)
241		    'CONFLICT 'NEED-MERGE))))
242	   (?J 'NEED-MERGE)		;not supported by standard CVS
243	   ((?U ?P)
244	    (if dont-change-disc 'NEED-UPDATE
245	      (cons 'UP-TO-DATE (if (eq code ?U) 'UPDATED 'PATCHED)))))
246	 path 'trust)))
247
248     (and
249      (cvs-match "pcl-cvs: descending directory \\(.*\\)$" (dir 1))
250      (setq cvs-current-subdir dir))
251
252     ;; A special cvs message
253     (and
254      (let ((case-fold-search t))
255	(cvs-match "cvs[.a-z]* [a-z]+: "))
256      (cvs-or
257
258       ;; CVS is descending a subdirectory
259       ;; (status says `examining' while update says `updating')
260       (and
261	(cvs-match "\\(Examining\\|Updating\\) \\(.*\\)$" (dir 2))
262	(let ((dir (if (string= "." dir) "" (file-name-as-directory dir))))
263	  (cvs-parsed-fileinfo 'DIRCHANGE "." dir)))
264
265       ;; [-n update] A new (or pruned) directory appeared but isn't traversed
266       (and
267	(cvs-match "New directory `\\(.*\\)' -- ignored$" (dir 1))
268	;; (cvs-parsed-fileinfo 'MESSAGE " " (file-name-as-directory dir))
269	;; These messages either correspond to a true new directory
270	;; that an update will bring in, or to a directory that's empty
271	;; on the current branch (either because it only exists in other
272	;; branches, or because it's been removed).
273	(if (ignore-errors
274	      (with-temp-buffer
275		(insert-file-contents (expand-file-name
276				       ".cvsignore" (file-name-directory dir)))
277		(goto-char (point-min))
278		(re-search-forward
279		 (concat "^" (regexp-quote (file-name-nondirectory dir)) "/$")
280		 nil t)))
281	    t		       ;The user requested to ignore those messages.
282	  (cvs-parsed-fileinfo '(NEED-UPDATE . NEW-DIR) dir t)))
283
284       ;; File removed, since it is removed (by third party) in repository.
285       (and
286	(cvs-or
287	 (cvs-match "warning: \\(.*\\) is not (any longer) pertinent$" (file 1))
288	 (cvs-match "`\\(.*\\)' is no longer in the repository$" (file 1))
289         (cvs-match "\\(.*\\) is no longer in the repository$" (file 1)))
290	(cvs-parsed-fileinfo
291	 (if dont-change-disc '(NEED-UPDATE . REMOVED) 'DEAD) file))
292
293       ;; [add]
294       (and
295	(cvs-or
296	 (cvs-match "scheduling file `\\(.*\\)' for addition.*$" (path 1))
297	 (cvs-match "re-adding file \\(.*\\) (in place of .*)$" (path 1)))
298	(cvs-parsed-fileinfo 'ADDED path))
299
300       ;; [add] this will also show up as a `U <file>'
301       (and
302	(cvs-match "\\(.*\\), version \\(.*\\), resurrected$"
303		   (path 1) (base-rev 2))
304	;; FIXME: resurrection only brings back the original version,
305	;; not the latest on the branch, so `up-to-date' is not always
306	;; what we want.
307	(cvs-parsed-fileinfo '(UP-TO-DATE . RESURRECTED) path nil
308			     :base-rev base-rev))
309
310       ;; [remove]
311       (and
312	(cvs-match "removed `\\(.*\\)'$" (path 1))
313	(cvs-parsed-fileinfo 'DEAD path))
314
315       ;; [remove,merge]
316       (and
317	(cvs-match "scheduling `\\(.*\\)' for removal$" (file 1))
318	(cvs-parsed-fileinfo 'REMOVED file))
319
320       ;; [update] File removed by you, but not cvs rm'd
321       (and
322	(cvs-match "warning: \\(.*\\) was lost$" (path 1))
323	(cvs-match (concat "U " (regexp-quote path) "$"))
324	(cvs-parsed-fileinfo (if dont-change-disc
325				 'MISSING
326			       '(UP-TO-DATE . UPDATED))
327			     path))
328
329       ;; Mode conflicts (rather than contents)
330       (and
331	(cvs-match "conflict: ")
332	(cvs-or
333	 (cvs-match "removed \\(.*\\) was modified by second party$"
334		    (path 1) (subtype 'REMOVED))
335	 (cvs-match "\\(.*\\) created independently by second party$"
336		    (path 1) (subtype 'ADDED))
337	 (cvs-match "\\(.*\\) is modified but no longer in the repository$"
338		    (path 1) (subtype 'MODIFIED)))
339	(cvs-match (concat "C " (regexp-quote path)))
340	(cvs-parsed-fileinfo (cons 'CONFLICT subtype) path))
341
342       ;; Messages that should be shown to the user
343       (and
344	(cvs-or
345	 (cvs-match "move away \\(.*\\); it is in the way$" (file 1))
346	 (cvs-match "warning: new-born \\(.*\\) has disappeared$" (file 1))
347	 (cvs-match "sticky tag .* for file `\\(.*\\)' is not a branch$"
348		    (file 1)))
349	(cvs-parsed-fileinfo 'MESSAGE file))
350
351       ;; File unknown.
352       (and (cvs-match "use `.+ add' to create an entry for \\(.*\\)$" (path 1))
353	    (cvs-parsed-fileinfo 'UNKNOWN path))
354
355       ;; [commit]
356       (and (cvs-match "Up-to-date check failed for `\\(.+\\)'$" (file 1))
357	    (cvs-parsed-fileinfo 'NEED-MERGE file))
358
359       ;; We use cvs-execute-multi-dir but cvs can't handle it
360       ;; Probably because the cvs-client can but the cvs-server can't
361       (and (cvs-match ".* files with '?/'? in their name.*$")
362	    (not cvs-execute-single-dir)
363	    (setq cvs-execute-single-dir t)
364	    (cvs-create-fileinfo
365	     'MESSAGE "" " "
366	     "*** Add (setq cvs-execute-single-dir t) to your .emacs ***
367	See the FAQ file or the variable's documentation for more info."))
368
369       ;; Cvs waits for a lock.  Ignored: already handled by the process filter
370       (cvs-match "\\[..:..:..\\] \\(waiting for\\|obtained\\) .*lock in .*$")
371       ;; File you removed still exists.  Ignore (will be noted as removed).
372       (cvs-match ".* should be removed and is still there$")
373       ;; just a note
374       (cvs-match "use ['`].+ commit' to \\sw+ th\\sw+ files? permanently$")
375       ;; [add,status] followed by a more complete status description anyway
376       (and (cvs-match "nothing known about \\(.*\\)$" (path 1))
377	    (cvs-parsed-fileinfo 'DEAD path 'trust))
378       ;; [update] problem with patch
379       (cvs-match "checksum failure after patch to .*; will refetch$")
380       (cvs-match "refetching unpatchable files$")
381       ;; [commit]
382       (cvs-match "Rebuilding administrative file database$")
383       ;; ???
384       (cvs-match "--> Using per-directory sticky tag `.*'")
385
386       ;; CVS is running a *info program.
387       (and
388	(cvs-match "Executing.*$")
389	;; Skip by any output the program may generate to stdout.
390	;; Note that pcl-cvs will get seriously confused if the
391	;; program prints anything to stderr.
392	(re-search-forward cvs-update-prog-output-skip-regexp))))
393
394     (and
395      (cvs-match "cvs[.ex]* \\[[a-z]+ aborted\\]:.*$")
396      (cvs-parsed-fileinfo 'MESSAGE ""))
397
398     ;; sadly you can't do much with these since the path is in the repository
399     (cvs-match "Directory .* added to the repository$")
400     )))
401
402
403(defun cvs-parse-merge ()
404  (let (path base-rev head-rev type)
405    ;; A merge (maybe with a conflict).
406    (and
407     (cvs-match "RCS file: .*$")
408     ;; Squirrel away info about the files that were retrieved for merging
409     (cvs-match "retrieving revision \\([0-9.]+\\)$" (base-rev 1))
410     (cvs-match "retrieving revision \\([0-9.]+\\)$" (head-rev 1))
411     (cvs-match "Merging differences between [0-9.]+ and [0-9.]+ into \\(.*\\)$"
412		(path 1))
413
414     ;; eat up potential conflict warnings
415     (cvs-or (cvs-match "\\(rcs\\)?merge:?\\( warning\\)?: \\(overlaps\\|conflicts\\) \\(or other problems \\)?during merge$" (type 'CONFLICT)) t)
416     (cvs-or
417      (and
418       (cvs-match "cvs[.ex]* [a-z]+: ")
419       (cvs-or
420	(cvs-match "conflicts found in \\(.*\\)$" (path 1) (type 'CONFLICT))
421	(cvs-match "could not merge .*$")
422	(cvs-match "restoring \\(.*\\) from backup file .*$" (path 1))))
423      t)
424
425     ;; Is it a succesful merge?
426     ;; Figure out result of merging (ie, was there a conflict?)
427     (let ((qfile (regexp-quote path)))
428       (cvs-or
429	;; Conflict
430	(and
431	 (cvs-match (concat "C \\(.*" qfile "\\)$") (path 1) (type 'CONFLICT))
432	 ;; C might be followed by a "suprious" U for non-mergeable files
433	 (cvs-or (cvs-match (concat "U \\(.*" qfile "\\)$")) t))
434	;; Successful merge
435	(cvs-match (concat "M \\(.*" qfile "\\)$") (path 1))
436	;; The file already contained the modifications
437	(cvs-match (concat "^\\(.*" qfile
438			   "\\) already contains the differences between .*$")
439		   (path 1) (type '(UP-TO-DATE . MERGED)))
440	t)
441       ;; FIXME: PATH might not be set yet.  Sometimes the only path
442       ;; information is in `RCS file: ...' (yuck!!).
443       (cvs-parsed-fileinfo (if dont-change-disc 'NEED-MERGE
444			      (or type '(MODIFIED . MERGED))) path nil
445			    :merge (cons base-rev head-rev))))))
446
447(defun cvs-parse-status ()
448  (let (nofile path base-rev head-rev type)
449    (and
450     (cvs-match
451      "===================================================================$")
452     (cvs-match "File: \\(no file \\)?\\(.*[^ \t]\\)[ \t]+Status: "
453		(nofile 1) (path 2))
454     (cvs-or
455      (cvs-match "Needs \\(Checkout\\|Patch\\)$"
456		 (type (if nofile 'MISSING 'NEED-UPDATE)))
457      (cvs-match "Up-to-date$"
458		 (type (if nofile '(UP-TO-DATE . REMOVED) 'UP-TO-DATE)))
459      (cvs-match "File had conflicts on merge$" (type 'MODIFIED))
460      (cvs-match ".*[Cc]onflict.*$"	(type 'CONFLICT))
461      (cvs-match "Locally Added$"	(type 'ADDED))
462      (cvs-match "Locally Removed$"	(type 'REMOVED))
463      (cvs-match "Locally Modified$"	(type 'MODIFIED))
464      (cvs-match "Needs Merge$"		(type 'NEED-MERGE))
465      (cvs-match "Entry Invalid"	(type '(NEED-MERGE . REMOVED)))
466      (cvs-match ".*$"			(type 'UNKNOWN)))
467     (cvs-match "$")
468     (cvs-or
469      (cvs-match " *Version:[ \t]*\\([0-9.]+\\).*$" (base-rev 1))
470      ;; NOTE: there's no date on the end of the following for server mode...
471      (cvs-match " *Working revision:[ \t]*-?\\([0-9.]+\\).*$" (base-rev 1))
472      ;; Let's not get all worked up if the format changes a bit
473      (cvs-match " *Working revision:.*$"))
474     (cvs-or
475      (cvs-match " *RCS Version:[ \t]*\\([0-9.]+\\)[ \t]*.*$" (head-rev 1))
476      (cvs-match " *Repository revision:[ \t]*\\([0-9.]+\\)[ \t]*\\(.*\\)$"
477		 (head-rev 1))
478      (cvs-match " *Repository revision:.*"))
479     (cvs-or (cvs-match " *Expansion option:.*") t)  ;Optional CVSNT thingie.
480     (cvs-or (cvs-match " *Commit Identifier:.*") t) ;Optional CVSNT thingie.
481     (cvs-or
482      (and ;; Sometimes those fields are missing.
483       (cvs-match " *Sticky Tag:[ \t]*\\(.*\\)$")      ; FIXME: use it.
484       (cvs-match " *Sticky Date:[ \t]*\\(.*\\)$")     ; FIXME: use it.
485       (cvs-match " *Sticky Options:[ \t]*\\(.*\\)$")) ; FIXME: use it.
486      t)
487     (cvs-or (cvs-match " *Merge From:.*") t) ;Optional CVSNT thingie.
488     (cvs-match "$")
489     ;; ignore the tags-listing in the case of `status -v'
490     (cvs-or (cvs-match " *Existing Tags:\n\\(\t.*\n\\)*$") t)
491     (cvs-parsed-fileinfo type path nil
492			  :base-rev base-rev
493			  :head-rev head-rev))))
494
495(defun cvs-parse-commit ()
496  (let (path file base-rev subtype)
497    (cvs-or
498
499     (and
500      (cvs-or
501       (cvs-match "\\(Checking in\\|Removing\\) \\(.*\\);$" (path 2))
502       t)
503      (cvs-match ".*,v  <--  \\(.*\\)$" (file 1))
504      (cvs-or
505       ;; deletion
506       (cvs-match "new revision: delete; previous revision: \\([0-9.]*\\)$"
507		  (subtype 'REMOVED) (base-rev 1))
508       ;; addition
509       (cvs-match "initial revision: \\([0-9.]*\\)$"
510		  (subtype 'ADDED) (base-rev 1))
511       ;; update
512       (cvs-match "new revision: \\([0-9.]*\\); previous revision: .*$"
513		  (subtype 'COMMITTED) (base-rev 1)))
514      (cvs-or (cvs-match "done$") t)
515      ;; In cvs-1.12.9 commit messages have been changed and became
516      ;; ambiguous.  More specifically, the `path' above is not given.
517      ;; We assume here that in future releases the corresponding info will
518      ;; be put into `file'.
519      (progn
520	;; Try to remove the temp files used by VC.
521	(vc-delete-automatic-version-backups (expand-file-name (or path file)))
522	;; it's important here not to rely on the default directory management
523	;; because `cvs commit' might begin by a series of Examining messages
524	;; so the processing of the actual checkin messages might begin with
525	;; a `current-dir' set to something different from ""
526	(cvs-parsed-fileinfo (cons 'UP-TO-DATE subtype)
527			     (or path file) 'trust
528			     :base-rev base-rev)))
529
530     ;; useless message added before the actual addition: ignored
531     (cvs-match "RCS file: .*\ndone$"))))
532
533
534(provide 'pcvs-parse)
535
536;; arch-tag: 35418375-1a23-40a0-957d-96b0262f91d6
537;;; pcvs-parse.el ends here
538