Demo Discussion
Forum Config Examples Contributions Vulnerabilities
  Discussion forum about ELOG, Page 406 of 807  Not logged in ELOG logo
New entries since:Thu Jan 1 01:00:00 1970
ID Date Icon Author Author Email Category OS ELOG Version Subject
  66567   Thu Oct 29 20:48:41 2009 Disagree David PilgramDavid.Pilgram@epost.org.ukBug reportLinux2.7.7-2252elog crashes with a long thread.
Hi Stefan,

I have a thread of 70 entries.  I added another entry, which was saved, but elog crashed.
It would restart, but crash every time I then tried to access that 71 entry thread.

By editing the yymmdda.log files to remove the latest entry, all was well again.
Add a test new entry (much smaller) also crashed elog as before.

If it is any help, this is the error message I caught on a console:

src/elogd.c:703: xrealloc: Assertion `*((unsigned int *) (temp + old_size)) == 0xdeadc0de' failed.
./log: line 1:  3123 Aborted    

Now I have got around this, by ending that thread with reference to a new one to continue, but is this to be
expected?  

If this is something (like memory allocation) that would have been in hiding from the start, I cannot imagine
that it is likely to be hit often enough to actually "bug fix" - it might, in any case, cause problems elsewhere.
  66566   Mon Oct 26 15:43:55 2009 Reply soren poulsensoren.poulsen@cern.chQuestionLinux2.7.7Re: Automatically generated incrementing tags (#)

soren poulsen wrote:

soren poulsen wrote:

Hi,

I am using the # character to generate automatically incrementing numbers for new messages.

My issue is that # is evaluated when you hit "New" but E-log is only aware of the new value being used when you hit "Submit".

So, two E-logs can have the same value substituted for # if two E-logs are being edited in parallel.

Maybe someone has a solution to this ?

Soren Poulsen

 

 

The solution is to use "Subst" instead of "Preset".

 This is not really resolved, since "Subst" creates a new number on both "New" and "Reply". I would like "Subst" to create a new number only on "New" and preserve this number through replies throughout the thread. I would like to be able to say "Subst thread = #" to make a new number for the thread and combine it with "Subst on reply thread = $thread" to preserve the number on replies, but this does not work. Maybe someone has already done this ?

Soren

  66565   Mon Oct 26 10:15:20 2009 Reply soren poulsensoren.poulsen@cern.chBug reportLinux2.7.7Re: User authorization file corruption

Stefan Ritt wrote:

soren poulsen wrote:

Hi,

Here is what happens (I think) if E-log encounters a full file system where it keeps the user authorization file:

1. When a user connects, E-log will make a backup of the file. The backup will be corrupt since the file system is full.

2. E-log will modify the contents of the original file, and write it back. The file will be corrupt since the file system is full.

3. Now, both the backup and the normal file are corrupt and you cannot log on, until someone cleans up the file system and restores a valid copy of the file.

Would it be possible to fix this ? Like abort if step 1 is not successful. And restore the backup file if step 2 is not successful.

Thanks a lot for you help 

Soren

Ok, I finally found some time (I'm pretty busy these days) to add a check for a potential full file system in SVN revision 2258. So before the password file would get corrupted, elog shows an error message about the full file system and just stops to work until space is freed up. 

Great. We fully appreciate that your are busy (with other things than E-log).

Thanks for the resolution.

Soren

  66564   Mon Oct 26 10:13:54 2009 Reply soren poulsensoren.poulsen@cern.chRequestLinux2.7.7Re: Option list length

Stefan Ritt wrote:

soren poulsen wrote:

Hi,

I use the following attribute definition:

Options <attribute> = <list> 

However, I am being limited by the list length limit of 100. I have 103 items, but I only see 100.

Could the limit be extended (to 200 for instance) ?

Thanks a lot for your help

Soren

You can change that yourself. Just find following line in elogd.c:

 

#define MAX_N_LIST      100

and change it to 200, then recompile. But you are there on your own, at some point you will get a stack overflow and elogd will crash, but I don't know exactly where this limit is.
 
Anyhow I would propose that if you have so many options in an attribute, that you better go and group these options somehow. Like using two attributes, where the first defines the group, and the second gets different list for each option of the first attribute using conditional attributes. Have a look here.

 

 

Thanks. This is a good explanation. It might indeed be better to re-group the options to have a shorter list.

Soren

  66563   Sat Oct 24 01:10:24 2009 Question Gabriele Sirrisirri@bo.infn.itBug reportAll2.7.7-2254"Collapse to last = 1" problem when reply twice to the same entry
Hello.

Please look at the entry 66525 of this forum (just 5 thread before this one):

 ->  chain.crt, posted by Gerhard Schneider on Thu Sep 3 21:55:52 2009         (66525)
  |->    Re: chain.crt, posted by Stefan Ritt on Fri Sep 4 08:33:16 2009       (66526)
  |->    Re: chain.crt, posted by Gerhard Schneider on Wed Oct 7 07:56:52 2009 (66556)

When you collapse the thread, it is collapsed to the 66526 instead of the 66556 (more recent)

  +      Re: chain.crt, posted by Stefan Ritt on Fri Sep 4 08:33:16 2009  

I guess it is because both 66526 and 66556 replies to the first entry. 
I have the same problem with Elog v2.7.7-2246 and Windows. 

In general, it seems to work well only if you always reply to the last entry of a thread.

  Thank you.


b.t.w. : is there any tip to always force reply to the last entry of a thread?
  66562   Wed Oct 21 19:06:59 2009 Reply Stefan Rittstefan.ritt@psi.chRequestAll2.7.7Re: feature req.: identify ELOG web pages via META element

Bill Pier wrote:

 

* Withdrawn *

The HTML layout produced by elogd is horrendous to deal with programmatically; I give up.

 


 

Hi,

I'm writing a greasemonkey script to slightly alter the look of the pages served by the ELOG server.  One difficulty that I'm struggling with is how to identify what type of page ELOG has created.  While I have several methods to determine the page type, such as a log entry vs. log entries summary, the solutions are not straight forward and not clean.  As far as I tell, there's no specific identification in HTML document currently that describes and identifies the type of page being served by the ELOG server.

So, I'm requesting that the pages created by ELOG be identified in some fashion with the META element, such as:

    <meta name="description" content="elog log entry" />

or

    <meta name="description" content="elog log summary" />

 

or even using the keywords attribute:

    <meta name="keywords" content="elog log summary" />

Sorry, I didn't yet have time to implement your request.

Have you considered to download an entry in plain text? Like this entry you can load with

https://midas.psi.ch/elogs/Forum/66555?cmd=Download

which should be much easier to interprete. If you write your entries in ELCode or plain format, it should be even easier. 

  66561   Fri Oct 16 12:21:45 2009 Reply Stefan Rittstefan.ritt@psi.chRequestLinux2.7.7Re: Option list length

soren poulsen wrote:

Hi,

I use the following attribute definition:

Options <attribute> = <list> 

However, I am being limited by the list length limit of 100. I have 103 items, but I only see 100.

Could the limit be extended (to 200 for instance) ?

Thanks a lot for your help

Soren

You can change that yourself. Just find following line in elogd.c:

 

#define MAX_N_LIST      100

and change it to 200, then recompile. But you are there on your own, at some point you will get a stack overflow and elogd will crash, but I don't know exactly where this limit is.
 
Anyhow I would propose that if you have so many options in an attribute, that you better go and group these options somehow. Like using two attributes, where the first defines the group, and the second gets different list for each option of the first attribute using conditional attributes. Have a look here.

 

 

  66560   Fri Oct 16 12:17:15 2009 Reply Stefan Rittstefan.ritt@psi.chBug reportLinux2.7.7Re: User authorization file corruption

soren poulsen wrote:

Hi,

Here is what happens (I think) if E-log encounters a full file system where it keeps the user authorization file:

1. When a user connects, E-log will make a backup of the file. The backup will be corrupt since the file system is full.

2. E-log will modify the contents of the original file, and write it back. The file will be corrupt since the file system is full.

3. Now, both the backup and the normal file are corrupt and you cannot log on, until someone cleans up the file system and restores a valid copy of the file.

Would it be possible to fix this ? Like abort if step 1 is not successful. And restore the backup file if step 2 is not successful.

Thanks a lot for you help 

Soren

Ok, I finally found some time (I'm pretty busy these days) to add a check for a potential full file system in SVN revision 2258. So before the password file would get corrupted, elog shows an error message about the full file system and just stops to work until space is freed up. 

ELOG V3.1.5-3fb85fa6