obsolete timestamp with cdr 1

[notuntoward]

When I run M-x ekg-capture, and then close the buffer with C-c C-c, I get the warning message:

obsolete timestamp with cdr 1

[ahyatt]

Thank you for reporting this bug. I don't think this error message is coming from ekg, and I haven't seen anything like it. Without further information, I'm not sure it's possible to proceed. What would help is if you could edebug into ekg-capture-finalize and see where in that function the message appears. Given that, I might be able to speculate on what could be going wrong.

[molekular]

I have the exact same message "obsolete timestamp with cdr 1" when I ekg-capture-finalize. I tried with emacs -q and only ekg and triples installed and get the same message.

toggle-debug-on-error gives no extra message at all, edebug-mode the same.

By the way, I also get the ewoc-locate: Wrong type argument: ewoc, nil error-message that was reported, #15.

I am an elisp-noob, but if I can do anything to help find the issue please advise.

[notuntoward]

I single-stepped through ekg-capture-finalize while watching the Messages window for the warning. I also wasn't able to identify the line of code where the warning originated, even though I stepped into called functions too. Then I ran out of time.

[ahyatt]

@notuntoward thank you for stepping through. Do you mean to say you got to the end without seeing the message? Did it come after the function exited? I think at this point don't bother stepping inside, just trace through and see what line it's coming from, which will help.

For both you are @molekular, can you try on an emacs run with -Q? Then load the triples library and emacsql-sqlite library, then ekg, and try again.

In addition, let me know if this happens every time, or only in certain circumstances.

[molekular]

I tried with emacs -Q: loaded the packages triples, emacsql-sqlite and ekg, loaded the respective libraries, ran ekg-capture and ekg-capture-finalize.

Message appears: "obsolete timestamp with cdr 1". At the same time the capture-buffer disappears and is replaced by the scratch-buffer. In the messages-buffer there is only "obsolete timestamp with cdr 1".

Yes, it happens every time, so nothing has been captured yet.

[ahyatt]

Looks like this is coming from decode_lisp_time, perhaps there's something interesting with your systems that is handling time in some strange way. What kind of system is this on, what is the value of the system-configuration variable?

Can you also run in a scratch buffer (time-convert (current-time) 'integer) and let me know what you get?

[ahyatt]

Also, can you try after evaluating (setq debug-on-message "timestamp")? That should cause you to enter the debugger when the message is printed, and then please paste the backtrace here.

[notuntoward]

@ahyatt, when I stepped through ekg-capture-finalize, the error message did show up in the messages window, but it was after I exited the function, so I couldn't figure out what caused it.

Here's that other information:

system-configuration is a variable defined in ‘C source code’.

Its value is "x86_64-w64-mingw32"

Value is string indicating configuration Emacs was built for.

Probably introduced at or before Emacs version 19.23.

When I run (time-convert (current-time) 'integer), I get:

1677710555 (#o14377752333, #x63ffd4db)

Doing (setq debug-on-message "timestamp") produced something more diagnostic, I hope:

Debugger entered--Lisp error: "obsolete timestamp with cdr 1"
  time-subtract((25599 54578 904559 0) (1677447982 . 1))
  #f(compiled-function (last-update) "Backup strategy to create a change daily at most.\nLAST-UPDATE is the time of the last update." #<bytecode -0x17c81e3a386fc3b0>)((1677447982 . 1))
  triples-backups-maybe-backup(#<emacsql-sqlite-connection emacsql-sqlite-connection-58e1b02728> nil)
  ekg-save-note(#s(ekg-note :id 33017693769 :text #("sadflkjadslfjs" 0 14 (fontified t wrap-prefix "")) :mode org-mode :tags ("date/2023-03-01") :creation-time 1677710638 :modified-time 1677710642 :properties nil))
  ekg--save-note-in-buffer()
  ekg-capture-finalize()
  funcall-interactively(ekg-capture-finalize)
  command-execute(ekg-capture-finalize)

[ahyatt]

Excellent! Even though I can't reproduce any failure, this helps narrow things down, and I found some weirdness, which I fixed. To test out, please evaluate the following on your system and try again; I hope it will be fixed. If so, I'll check it in and push this fix to the triples module.

(defun triples-backups-maybe-backup (db &optional filename)
  "If it is time for DB to be backed up, then back it up.
FILENAME is optional, as in `triples-connect', if not given will
default to the standard triple database given in
`triples-default-database-filename'."
  (let* ((backup-info (triples-backups-configuration db))
         (strategy-func (intern (format "triples-backups-strategy-%s"
                                        (plist-get backup-info :strategy)))))
    (unless backup-info
      (error "`triples-backups-setup' needs to be called on this database before trying to back up."))
    (unless (fboundp strategy-func)
      (display-warning
       'triples
       (format "Triples backup strategy %s not found, defaulting to `triples-backups-strategy-daily'"
               strategy-func)
       :error))
    (when (funcall (or (symbol-function strategy-func) #'triples-backups-strategy-daily)
                   (plist-get backup-info :last-update-time))
      (triples-backup db filename (plist-get backup-info :num-to-keep))
      (apply #'triples-set-type db 'database 'backup (plist-put backup-info :last-update-time (time-convert (current-time) 'integer))))))

(defun triples-backups-strategy-daily (last-update)
  "Backup strategy to create a change daily at most.
LAST-UPDATE is the time of the last update."
  (>= (/ (- (float-time (current-time)) (float-time last-update)) 86400)
      1))

[notuntoward]

That fixed it. I don't see the timestamp error message anymore. Thanks!

[ahyatt]

Awesome, thank you for your help in debugging this! I'll commit the fix today and the triples library should be released as a new version in GNU ELPA tomorrow. Nothing needs to be done on the ekg side.

[monnier]

The "obsolete timestamp with cdr" warning was added to Emacs-27 to try and catch uses of an old (HI . LO) timestamp representation (which clashes with the new (TIME . HZ) representation). It clearly misfires here. It was removed in Emacs-28.

[ahyatt]

Thank you for the explanation, @monnier!

[molekular]

Thank you so much, this fixed it for me too.