Error message could be improved - Specified time is not representable

Error message could be improved - Specified time is not representable

[Julian Bean]

Hi all,

On upgrading to emacs 23.3 (from emacs 23.1) my daily/weekly agenda became unusable, giving the error message "Specified time is not representable".

After a bit of poking around in the backtrace I tracked this down to an erroneous SCHEDULED date of 1st January 1904 buried inside a task. (This date was created by a bug in a script I was using to import TODOs from another system).

Evidently, for some reason, calling (encode-time 0 0 0 1 1 1904) on 23.3 causes the error above, whereas on 23.1 it's silently ignored?

I think the best thing for org-mode is to catch this error and report it to the user in a more understandable way so they can locate and fix the buggy timestamp - at least, include the full text of the timestamp so they can easily grep for it?

Jules

Re: Error message could be improved - Specified time is not representable

[Giovanni Ridolfi]

Julian Bean <jules@jellybean.co.uk> writes:

Hi Jules,

>
> On upgrading to emacs 23.3 (from emacs 23.1) my daily/weekly agenda
> became unusable, giving the error message "Specified time is not
> representable".

>
> [...] I tracked this down to an erroneous SCHEDULED date of 1st
> January 1904 buried inside a task. [...]
>
> Evidently, for some reason, calling (encode-time 0 0 0 1 1 1904) on
> 23.3 causes the error above, 

" This is probably related to the same problem as discussed in this
  thread:
  http://thread.gmane.org/gmane.emacs.orgmode/39206

  on the limitations of unix time (whether on Unix or not...)."

> whereas on 23.1 it's silently ignored?

Since it is an Emacs problem I think you should ask Emacs's 
developers:
emacs-devel@gnu.org 

or file a bug report: 
M-x report-emacs-bug or write to
bug-gnu-emacs@gnu.org

cheers,
Giovanni

Re: Error message could be improved - Specified time is not representable

[Julian Bean]

On 24 May 2011, at 09:16, Giovanni Ridolfi wrote:

> Julian Bean <jules@jellybean.co.uk> writes:
> 
>> 
>> [...] I tracked this down to an erroneous SCHEDULED date of 1st
>> January 1904 buried inside a task. [...]
>> 
>> Evidently, for some reason, calling (encode-time 0 0 0 1 1 1904) on
>> 23.3 causes the error above, 
> 
> " This is probably related to the same problem as discussed in this
>  thread:
>  http://thread.gmane.org/gmane.emacs.orgmode/39206
> 
>  on the limitations of unix time (whether on Unix or not...)."

Thanks, I missed that. You're right, that describes another manifestation of the same issue.

> 
>> whereas on 23.1 it's silently ignored?
> 
> Since it is an Emacs problem I think you should ask Emacs's 
> developers:
> emacs-devel@gnu.org 
> 
> or file a bug report: 
> M-x report-emacs-bug or write to
> bug-gnu-emacs@gnu.org

I disagree. It's not a clear emacs bug - the docstring for encode-time says very clearly "Years before 1970 are not guaranteed to work.  On some systems, year values as low as 1901 do work.". encode-time is working as documented and I certainly don't understand emacs' date-time internals well enough to suggest a better way.

The *org-mode* bug is, simply, the error message (hence my message title). If this exception does occur, it would be nice if org-mode would catch it, and provide better information to the user about which timestamp he needs to fix.

Jules