> > With elog it is possible to submit messages to a password protected
> > logbook without specifying the -u option.  I.e. NO PASSWORD is
> > necessary to submit a message.  I assume it is related to the problem
> > of expiring password-cookies while entering the message using a web
> > browser.
> 
> Indeed this problem is related to the expiring password cookies. As a 
> reminder: For the submission of a new entry, the password is checked when
one 
> presses the "New" button, but NOT for the "submit". This is because a 
> password can expire between the "New" and the "Submit", so a entered message 
> could not be sent. The question is now what to do with the standalone
"elog".
> 
> Right now, elog does a normal submission where the password is not checked, 
> which is maybe not what one wants. But what to do? If elog sends a special 
> flag "please do check password on submit", someone could analyze the source 
> code, remove the flag from elog and then still submit messages without a 
> password. If I put an additional flag to the web browser submission "please 
> do not check the password since the cookie might have been expired", someone 
> can add this flag into elog and still bypass the password checking.

I guess it cannot and doesn't have to be 100% save.  Maybe if the web
interface is used for a new message a long random number (let's call
it newID) can be included, which elog remembers for some time (say 1
day).  Now elogd accepts a new message only if 

  1) the cookies is there and valid or
  2) if the cookies are NOT THERE, but the newID matches one of the
       stored ones.     

The new message is rejected if the cookies are there, but are wrong.

> Anothe thing which bothers me is if you specify the password explicitly on 
> the command line of elog, it's visible in some scripts etc, which yould be a 
> security issue as well.

Maybe the encoded password should be specified.  I use wget to
retrieve some entries automatically over a cron job and with wget
you specify a cookie-file with --cookie-file (or something like
this).  The content of this file corresponds to the content of the
netscape cookie file.

> 
> Any ideas?

Can one delete or edit messages with elog?  If yes then this should not be
possible.

> I guess it cannot and doesn't have to be 100% save.  Maybe if the web
> interface is used for a new message a long random number (let's call
> it newID) can be included, which elog remembers for some time (say 1
> day).  Now elogd accepts a new message only if 
> 
>   1) the cookies is there and valid or
>   2) if the cookies are NOT THERE, but the newID matches one of the
>        stored ones.     
> 
> The new message is rejected if the cookies are there, but are wrong.

Ok that sounds a good idea to me, I will work on that.

> Can one delete or edit messages with elog?  If yes then this should not be
> possible.

No this is not possible.

> If a message is submitted via the elog command then the reply string '> ' 
is only inserted in the first line if a reply is made.
> This message was submitted with the following commandline:
> elog -v -h midas.psi.ch -p 80 -s elogdemo -l Forum -a "Icon=icon4.gif" -
a "Author=Heiko Scheit"  -a "Author Email=h.scheit@mpi-hd.mpg.de" -
a "Subject=elog submit problem" "...Message-text..."
> Please press the reply button to see the problem.

Has been fixed in 2.0.4 (as you can see from this entry). Please make sure 
not to submit too long lines via "elog". For browser submissions, the line 
length is limited to 78 characters, but not for "elog" submissions.

> I get the following compiler warnings about sprintf() buffer overflows. I suggest sprintf() should be replaced by std::string msprintf() from 
> midas. K.O.

I started to convert some sprintf() to snprintf(), but I still have 824 cases to go... Ideally, all should be converted to std::string. Will be some job for my retirement ;-)

Stefan

Have you tried to restart the elogd server? The CLOSE_WAIT could be dangling network connections, which were not properly closed by the browser.

We do run ELOG on a Linux server and see about weekly crashes, too. It seems to be connected to the authentication process (Kerberos, File), but we could not nail it down yet.

But we have set up a supervision process that checks every minute if the "elogd" process is still running. If not, it simply restarts it.

If ELOG is down for two minutes a week, this is fine for our users.

Dear Stefano,

Try the entry I wrote some time ago elog:68655

David.


> Dear expert,
>   I'm running the latest git version of elog ELOG V3.1.4-395e101a on ubuntu 20.04.2.
> I'm experiencing frequent hangs of the elog server: the status is always reported as running, but the web server is not responding.
> The only hint I have of something strange is that the elogd process is using a lot of CPU (50-100%), the log do not show anything suspect 
> as far as I can see.
> 
> Has anyone experienced something similar or has any idea how can I start to debug the problem?
> 
> Sorry for lack of many information, but I don't know what to look at.
> 
> Thanks in advance
>   Stefano

Usually a restart of the elogd server helps. If the problem persists, one of the logbooks might be corrupt. Try to disable one logbook at a time to figure out which one it is. Then 
remove that one and set it up freshly.

Stefan