[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
bug#71223: 30.0.50; stack overflow after very fast opening and closing o
From: |
Stefan Monnier |
Subject: |
bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames |
Date: |
Wed, 29 May 2024 17:07:34 -0400 |
User-agent: |
Gnus/5.13 (Gnus v5.13) |
[ Hi Chong, see below a question for you. ]
> Could you try the patch below which sequentializes the processing of the
> process-filters to avoid such recursion?
Here's a new version of the patch which fixes various shortcomings
I noticed while trying it out. Also it shortens the wait times after
displaying a message by applying the `sit-for` principle of not waiting
if there's some "pending input" (where I consider `emacsclient`
processes to be pending inputs as well). This way, the `sit-for`
waiting times don't accumulate so much.
Also I noticed a weird thing in
commit 2a847524ab57b1b3d6eaa7e12b96be52dbb79509
Author: Chong Yidong <cyd@gnu.org>
Date: Sat Oct 2 20:03:44 2010 -0400
* lisp/server.el (server-process-filter, server-return-error): Give
emacsclient time to shut down after receiving an error string.
* etc/NEWS: Document tweak to emacsclient exit status.
[...]
diff --git a/lisp/server.el b/lisp/server.el
index 0f1b0219a2c..e661f055e1a 100644
--- a/lisp/server.el
+++ b/lisp/server.el
@@ -876,6 +876,9 @@ server-process-filter
(server-log "Authentication failed" proc)
(server-send-string
proc (concat "-error " (server-quote-arg "Authentication failed")))
+ ;; Before calling `delete-process', give emacsclient time to
+ ;; receive the error string and shut down on its own.
+ (sit-for 1)
(delete-process proc)
;; We return immediately
(return-from server-process-filter)))
@@ -1129,6 +1132,9 @@ server-return-error
proc (concat "-error " (server-quote-arg
(error-message-string err))))
(server-log (error-message-string err) proc)
+ ;; Before calling `delete-process', give emacsclient time to
+ ;; receive the error string and shut down on its own.
+ (sit-for 5)
(delete-process proc)))
(defun server-goto-line-column (line-col)
Why do we wait for 5s in one spot but 1s in another?
Stefan
diff --git a/lisp/server.el b/lisp/server.el
index b65053267a6..06bba74bfd7 100644
--- a/lisp/server.el
+++ b/lisp/server.el
@@ -438,7 +438,8 @@ server-sentinel
(ignore-errors
(delete-file (process-get proc :server-file))))
(server-log (format "Status changed to %s: %s"
- (process-status proc) msg) proc)
+ (process-status proc) msg)
+ proc)
(server-delete-client proc))
(defun server--on-display-p (frame display)
@@ -1046,6 +1047,9 @@ server-execute-continuation
(process-put proc 'continuation nil)
(if continuation (ignore-errors (funcall continuation)))))
+(defvar server--pending-filter-calls nil)
+(defvar server--processing-pending-filters nil)
+
(cl-defun server-process-filter (proc string)
"Process a request from the server to edit some files.
PROC is the server process. STRING consists of a sequence of
@@ -1145,6 +1149,37 @@ server-process-filter
`-suspend'
Suspend this terminal, i.e., stop the client process.
Sent when the user presses \\[suspend-frame]."
+ ;; Push this to the end of the list, so the list is in the order in which
+ ;; we need to process it.
+ ;; FIXME: This implies an O(N²) worst-case, which is not good:
+ ;; we should use a "true" O(N) queue.
+ (setq server--pending-filter-calls
+ (nconc server--pending-filter-calls (list (cons proc string))))
+ (unless server--processing-pending-filters
+ (server-process-pending-filters)))
+
+(defun server-process-pending-filters ()
+ (let ((server--processing-pending-filters t))
+ (unwind-protect
+ (while server--pending-filter-calls
+ (let* ((oldest (pop server--pending-filter-calls)))
+ (server--process-filter (car oldest) (cdr oldest))))
+ ;; In case we're exiting early (e.g. for `server-goto-toplevel'),
+ ;; make sure we continue running the other pending filters.
+ (when server--pending-filter-calls
+ (run-with-timer 0 nil #'server-process-pending-filters)))))
+
+(defun server--message-timed (time &rest args)
+ (apply #'message args)
+ ;; FIXME: Ideally we should not need `sit-for' here and instead use
+ ;; some `message-timed' call which makes sure the message is visible
+ ;; without forcing us to wait here.
+ ;; If there's already another process-filter pending, skip `sit-for',
+ ;; just as it does when there's pending user input.
+ (unless (consp server--pending-filter-calls)
+ (sit-for time)))
+
+(cl-defun server--process-filter (proc string)
(server-log (concat "Received " string) proc)
;; First things first: let's check the authentication
(unless (process-get proc :authenticated)
@@ -1158,8 +1193,7 @@ server-process-filter
;; Display the error as a message and give the user time to see
;; it, in case the error written by emacsclient to stderr is not
;; visible for some reason.
- (message "Authentication failed")
- (sit-for 2)
+ (server--message-timed 2 "Authentication failed")
(server-send-string
proc (concat "-error " (server-quote-arg "Authentication failed")))
(unless (eq system-type 'windows-nt)
@@ -1169,10 +1203,10 @@ server-process-filter
(delete-terminal terminal))))
;; Before calling `delete-process', give emacsclient time to
;; receive the error string and shut down on its own.
- (sit-for 1)
- (delete-process proc)
+ ;; FIXME: Why do we wait 1s here but 5s in the other one?
+ (run-with-timer 1 nil #'delete-process proc)
;; We return immediately.
- (cl-return-from server-process-filter)))
+ (cl-return-from server--process-filter)))
(let ((prev (process-get proc 'previous-string)))
(when prev
(setq string (concat prev string))
@@ -1507,8 +1541,7 @@ server-return-error
;; Display the error as a message and give the user time to see
;; it, in case the error written by emacsclient to stderr is not
;; visible for some reason.
- (message (error-message-string err))
- (sit-for 2)
+ (server--message-timed 2 (error-message-string err))
(server-send-string
proc (concat "-error " (server-quote-arg
(error-message-string err))))
@@ -1520,8 +1553,8 @@ server-return-error
(delete-terminal terminal))))
;; Before calling `delete-process', give emacsclient time to
;; receive the error string and shut down on its own.
- (sit-for 5)
- (delete-process proc)))
+ ;; FIXME: Why do we wait 5s here but 1s in the other one?
+ (run-with-timer 5 nil #'delete-process proc)))
(defun server-goto-line-column (line-col)
"Move point to the position indicated in LINE-COL.
bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Stefan Monnier, 2024/05/28
bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames,
Stefan Monnier <=
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Eli Zaretskii, 2024/05/30
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Stefan Monnier, 2024/05/30
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Eli Zaretskii, 2024/05/30
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Stefan Monnier, 2024/05/30
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Stefan Monnier, 2024/05/30
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Eli Zaretskii, 2024/05/31
- bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Stefan Monnier, 2024/05/31
bug#71223: 30.0.50; stack overflow after very fast opening and closing of frames, Daniel Clemente, 2024/05/30