Using Elog 2.9.2

Elog crashes when making 10 replies, I narrowed the crash down to the Propagate Attributes setting.

I have an attribute "Status" that can be toggled between "Open" and "Closed" and that propagates all replies.  On the 10th reply the application crashed, this is repeatable 100% of the time.  Without the propagate option everything works fine.

Attached is my config file parsed down.

Is there a way around this, or is there a way on reply to change the attribute of the log note you replied to from open to closed without using the propagate option?

Hi Stephen,

I see that you don't allow branching in your threads.  Why do you need to propagate the Status throughtout the thread?  Why not just mark the latest entry and (incase it is necessary) 'collapse to last = 1'

I'm not saying that the bug (if bug it is, rather than a preset limitation) you've found should not be fixed, but I'm puzzled as to why you happen to use the feature.  I can see the point if an initial entry provides a whole tree of branches (and a limit of 10 is rather limiting).  OK, I know if it is historic, it cannot easily be changed because of other users.  Or if some users will primarily be responding to emails rather viewing the logbook via a browser.  I'm a single (ab)user elog system myself, so I'm very tolerant of changing how elog works if it offers an improvement.

 Thank you for the reply, the not allowing branching was added during the troubleshooting phase of the crashing.  Initially I thought the branches caused the issue because crashing only occured on log notes that branched as I only saw the crash on notes that branched, this proved not to be the case.  Even with branches the system will crash once there are 10 replies in the note with propagate on.

As for use case, with  a few users I needed a way to represent a note as open so that the next person knew work still needed to be done.  Once work was completed all notes close, the reason for not using edit is so I have a clear record of what happened (although I may need to change to edit if 10 is the max).  I will experiment with the collapse to last to see how it looks.  Any other suggestions would be helpful.

Bad news: I cannot reproduce your problem with the latest elog version 2.9.2 (e7ba466) on scientific linux 6.0

I had to remove some lines in your config, since I don't have your password file.

I'll attach the config file, please have a look if it crashes on your server with the latest elogd version.

Andreas

 Thanks for the reply, I tried the very basic cfg and the second one you offered.  Running 2.9.2-2475 on a Windows 2008 R2 V.Server, I still crashed on attempt 10 never fail.  Windows detects the failure and gives this message:

Faulting application name: elogd.exe, version: 0.0.0.0, time stamp: 0x51248707

Faulting module name: elogd.exe, version: 0.0.0.0, time stamp: 0x51248707

Exception code: 0xc00000fd

Fault offset: 0x00065127

Faulting process id: 0x340

Faulting application start time: 0x01ceefa87fea0aea

Faulting application path: D:\ELOG\elogd.exe

Faulting module path: D:\ELOG\elogd.exe

Report Id: ceccc3da-5b9b-11e3-80f1-5ac95c924b0b

It only happens when propagate is on, I have to be missing some security setting in Windows maybe?  Don't know if that helped at all, but I'm out of ideas. 

 I have been able to compile Elogd successfully in Windows with Visual Studio 2010 but not with Elog.  Attached are errors, it complaint about buffer[i] type char * being assigned type void *.  If anyone has been successful compile the elog, please give me a hint.  Thanks.

Just put (char *) in front of all malloc(). I will include the fix in the next official version. 

/Stefan

You don't need "elog" to run the logbook, the executable "elog" is only needed to create entries without the web interface. Stefan should adapt "elog.c" some day to be compatible with the default switches of modern, paranoid compilers; but for the moment you can ignore this.

If you use the newly created "elogd", does that reproduce the former problem?

 I don't see it has crashed when testing your cfg.  However, one thing I must do is to comment out the alarm_handler function in order to compile the elogd successfully.  Otherwise, it shows an undefined alarm, see in the 2nd attachment.  Although it did not crash as tested with your cfg, I have experienced the new version sometimes crashed on our system but then it self restarted (different scenario since my cfg does not use Propagate.  I have not been able to narrow down yet.

Yes, the alarm() function is wrong here. I removed it from the Windows version and committed the code to GIT. 

Was a tricky problem. When you use "Propagate Attributes", a routine is called inside elogd which recursively goes through all replies. It uses quite some memory for temporary storage of attributes and attachments. So after a certain time you get a stack overflow since you run out of memory. Since the stack size is different under different operating systems, it was hard for me to reproduce it initially. Now I use some different memory (dynamic heap instead stack) which should fix the problem. Can you pull from GIT and give it a try?

/Stefan 

 Unfortunately, I am a little outside of my element when trying to compile it.  I will ask around and see if someone here could give me a hand.  Thanks for looking into this for me.

PS.  If anyone else has managed to compile this could you give me a hand =)

 This resolved the issue, thanks for the help.  I have tested it on server machines and was able to go over 10 each time.

Thank you for resolving this issue for me.

 I was able to manage and compile the latest code from GIT.  It runs fine so far.  Just a hint, in order to compile successfully, there are some steps that you may need to modify your header file depend on how you store your mxml, krb5, OpenSSL files and directories.

Hello community,

I have a feature request for the form of the ELOG.

For some users, especially coming from a MAC background, it is very inconvenient to press enter (ie. on the subject field) and post a message (or get prompted to do so). Instead, enter in these fields should work as a tab.

Cheers, keep up the good work,

Daniel

Have a look at this forum. If you press <enter> on the subject field, you get asked "do you really want to submit this entry" and you have the chance to cancel and continue editing. You should get the same with the current elog version. 

Hi all,

I am successfully using ELOG V2.9.2-2455 to manage several laboratory logs.

I am looking for a way, if possible, to format the subject attribute in bold type while in thread display.

I've used the following code

Display Subject = <b>$subject</b>

with no success, then I've tried

Subst Subject = <b>$subject</b>

again with no success. It seems that the <b> tag is not iterpreted.

Have you any suggestion about this?

Thank you in advance.

Paolo

1st: If you want to use HTML tags in the display of attributes, you need to add "Allow HTML = 1" in the configuration.

2nd: The proper command is not "Display <attribute> = ..." but "Change <attribute> = ...". It is actually correct in the documentation (Change <attribute> = <string>), but the example is wrong (Display Telephone = <a href="http://any.company.com/telbook.cgi?search=$Name">$Name's telephone number</a>)

Stefan should fix the example in https://midas.psi.ch/elog/config.html#attrib some day.

Thanks, I fixed the documentation.

/Stefan 

Dear Andreas,

thank you very much for your prompt reply and for you suggestions, which helped me to solve the problem. Now it's all working perfectly.

Cheers,

Paolo

 Hi All,

I'm looking for a way (or if it is even possible?) for users to browse between logbooks without being logged out of their current logbook if they click on a logbook they are not authorised to access?
If they accidentally choose the wrong logbook they are forced to log back in again?

No guest access has been configured on any log.

TIA
 

When compiling elog on OSX 10.9 (Mavericks), I get the error below.

Elog will compile without error if I add -D_FORTIFY_SOURCE=0 to CFLAGS in Makefile, but I'm not sure whether this is a good idea.

$ make

cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall  -I../mxml  -DHAVE_SSL -w -c -o crypt.o src/crypt.c

cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall  -I../mxml  -DHAVE_SSL -o elog src/elog.c crypt.o -lssl

src/elog.c:125:8: error: expected parameter declarator

size_t strlcpy(char *dst, const char *src, size_t size)

       ^

/usr/include/secure/_string.h:105:44: note: expanded from macro 'strlcpy'

  __builtin___strlcpy_chk (dest, src, len, __darwin_obsz (dest))

                                           ^

/usr/include/secure/_common.h:39:62: note: expanded from macro '__darwin_obsz'

#define __darwin_obsz(object) __builtin_object_size (object, _USE_FORTIFY_LEVEL > 1 ? 1 : 0)

                                                             ^

/usr/include/secure/_common.h:30:32: note: expanded from macro '_USE_FORTIFY_LEVEL'

#    define _USE_FORTIFY_LEVEL 2

                               ^

src/elog.c:125:8: error: expected ')'

/usr/include/secure/_string.h:105:44: note: expanded from macro 'strlcpy'

  __builtin___strlcpy_chk (dest, src, len, __darwin_obsz (dest))

                                           ^

/usr/include/secure/_common.h:39:62: note: expanded from macro '__darwin_obsz'

#define __darwin_obsz(object) __builtin_object_size (object, _USE_FORTIFY_LEVEL > 1 ? 1 : 0)

                                                             ^

/usr/include/secure/_common.h:30:32: note: expanded from macro '_USE_FORTIFY_LEVEL'

#    define _USE_FORTIFY_LEVEL 2

                               ^

src/elog.c:125:8: note: to match this '('

/usr/include/secure/_string.h:105:44: note: expanded from macro 'strlcpy'

  __builtin___strlcpy_chk (dest, src, len, __darwin_obsz (dest))

                                           ^

/usr/include/secure/_common.h:39:53: note: expanded from macro '__darwin_obsz'

#define __darwin_obsz(object) __builtin_object_size (object, _USE_FORTIFY_LEVEL > 1 ? 1 : 0)

                                                    ^

All over sudden gcc comes with its own version of "strlcpy", which I had defined "manually" since many years inside ELOG. Using -DFORTIFY_SOURCE=0 will not harm, so you can use it. The "real" solution is to take our ELOG's strlcpy/strlcat, which I did on the current SVN version.

Best regards,
Stefan 

Ok, I tried updating my SVN working copy, but I didn't get any updates past elog rev. 2494, mxml rev. 74.  I undid my changes to Makefile, tried to compile, but got the same errors.  

I then pulled down elog and mxml with git, and these are working for me with no errors.  Thanks!

SVN is obsolete and will NOT be maintained any more, since we completely switched to GIT. Actually I will disable the service soon. 

 I too am having issues installing on Mac OS X 10.9.1. I changed CFLAGS in makefile to;

CFLAGS += -O3 -funroll-loops -fomit-frame-pointer -W -Wall -D_FORTIFY_SOURCE=0

I get many lines of errors like this when I run make;

/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.9.sdk/usr/include/openssl/ssl.h:1491:6: note: 'SSL_accept' declared here

int     SSL_accept(SSL *ssl) DEPRECATED_IN_MAC_OS_X_VERSION_10_7_AND_LATER;

        ^

src/elogd.c:28809:19: warning: 'SSL_set_fd' is deprecated: first deprecated in OS X 10.7 [-Wdeprecated-declarations]

                  SSL_set_fd(ka_ssl_con[i_min], ka_sock[i_min]);

                  ^

If you would use the Makefile from the GIT repository there would be no errors under OS X 10.9.1:

/elog$ make                                                                                                                                                                 
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -w -c -o crypt.o src/crypt.c                                         
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -o elog src/elog.c crypt.o -lssl                                     
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -w -c -o regex.o src/regex.c                                         
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -w -c -o auth.o src/auth.c                                           
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -c -o mxml.o ../mxml/mxml.c                                          
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -c -o strlcpy.o ../mxml/strlcpy.c                                    
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -o elogd src/elogd.c crypt.o auth.o regex.o mxml.o strlcpy.o -lssl   
cc -O3 -funroll-loops -fomit-frame-pointer -W -Wall -Wno-deprecated-declarations -I../mxml  -DHAVE_SSL -o elconv src/elconv.c -lssl                                         
/elog$                                                                                                                                                                      

Actually what you need is -Wno-deprecated-declarations to suppress the warnings. The open SSL functions will at some point be removed from OSX, they have their own implementation of SSL. So then we either have to ship openSSL together with elog or use Apple's implementatoin. But for now we are still fine.

/Stefan

In the past couple of days, I seem to have developed a problem with making entries into elog's displayed with Chrome (the latest, 31.0.1650.63) on 
Mac OS (10.9.1, the latest).  The problem occurs with attempting to edit or enter HTML encoded pages with fckedit; although pages render correctly 
in list mode, if you try to edit or enter an entry, the page is blank, the cursor is missing, you can't see text or type new text.  I ran elogd -v by hand, 
and there were no obvious problems, and I looked at the developer consoles in Chrome, and the only place I see any hint of what the problem might 
be is the Javascript console which says this:

event.returnValue is deprecated. Please use the standard event.preventDefault() instead.
Uncaught SecurityError: Blocked a frame with origin "http://localhost:8080" from accessing a frame with origin "chrome-
extension://pioclpoplcdbaefihamjohnefbikjilc".  The frame requesting access has a protocol of "http", the frame being accessed has a protocol of 
"chrome-extension". Protocols must match.
 fckeditorcode_gecko.js:36

It works ok in Safari, but it would be nice to use Chrome, and it was working ok until recently.  I don't think the problem occurred when I updated to 
Mac OS 10.9.1, but I don't keep careful track of the Chrome version.  It's not critical, but I pretty much exhausted what I knew how to debug.  I have 
close to the latest elog (2.9.2-2455), although I see the same phenomenon on this elog (.2.9.2-2475) and I think it's related to this thread:

http://productforums.google.com/forum/#!msg/maps/hQhwWA56NbA/2XL35dU7le4J

I tried the prescription in the October 22 entry, but it didn't seem to help, although I wasn't sure I had really tested it with compressed javascript and 
cache and what have you.

> In the past couple of days, I seem to have developed a problem with making entries into elog's displayed with Chrome (the latest, 31.0.1650.63) on 
> Mac OS (10.9.1, the latest).  The problem occurs with attempting to edit or enter HTML encoded pages with fckedit; although pages render correctly 
> in list mode, if you try to edit or enter an entry, the page is blank, the cursor is missing, you can't see text or type new text. 

That's strange. I just tried myself (by accident I have the same versions of Chrome and OSX) and it just worked fine. The fckedit code has not change a long time, so I guess it's not related to the exact version of 
elogd. Anyhow I want to switch to ckedit when I get some time, which maybe fixes the problem. What happens if you try to write in this forum, do you have the same problem? Sometime the fcdedit code take 
quite long time to load when accessed remotely. If your browser gives up, you might hat to click on "reload". In chrome there is also the "Developer tools" window, which shows you all HTTP requests and 
responses on the network. Otherwise I run out of ideas what could be different for you compared to me.

Cheers,
Stefan

> > In the past couple of days, I seem to have developed a problem with making entries into elog's displayed with Chrome (the latest, 31.0.1650.63) on 
> > Mac OS (10.9.1, the latest).  The problem occurs with attempting to edit or enter HTML encoded pages with fckedit; although pages render correctly 
> > in list mode, if you try to edit or enter an entry, the page is blank, the cursor is missing, you can't see text or type new text. 
> 
> That's strange. I just tried myself (by accident I have the same versions of Chrome and OSX) and it just worked fine. The fckedit code has not change a long time, so I guess it's not related to the exact version of 
> elogd. Anyhow I want to switch to ckedit when I get some time, which maybe fixes the problem. What happens if you try to write in this forum, do you have the same problem? Sometime the fcdedit code take 
> quite long time to load when accessed remotely. If your browser gives up, you might hat to click on "reload". In chrome there is also the "Developer tools" window, which shows you all HTTP requests and 
> responses on the network. Otherwise I run out of ideas what could be different for you compared to me.
> 
> Cheers,
> Stefan

Later on, I tried the prescription pointed to in that thread that says in scripts/fckeditor/editor/js/fckeditorcode_gecko.js to replace

FCKTools.FixDocumentParentWindow = function(A){ if (A.document) A.document.parentWindow=A; for (var i=0;i<A.frames.length;i++) FCKTools.FixDocumentParentWindow(A.frames[i]);};

with: 

FCKTools.FixDocumentParentWindow = function(A){try{ if (A.document) A.document.parentWindow=A;} catch(e){};for (var i=0;i<A.frames.length;i++) FCKTools.FixDocumentParentWindow(A.frames[i]);};

and after I got all my brackets straight, my pages were visible.  I'm still not sure when this turned up, since I use my elog every day (hour) and so I'd hardly miss it when I upgraded various things, although I don't watch Chrome updating itself that closely.

Have a good holiday!

 Hello,

I have followed the very simple steps on the Download page for checking out and compiling from GIT.  That works perfect and there is no issue.  

The problem I have is-- it is not clear to me where to put the 'elog' dir that I have after I 'make' and 'make install'.  Or, is there an installer script afterwards that I run? I'm installing on Debian 7 and trying to upgrade from 2.5.2 (which was installed using apt-get.)

Hi Ryan,

as far as I remember the Debian package is not supported any more. The "make install" assumes Red-hat style installation directories (you can see it in elog/Makefile, all the installation directories are installed there).

I have no idea where Debian is supposed to install the binaries. But you should be able to use GNU "locate" to find the old files: "locate elog" and "locate elogd" should tell you where the old binaries had been installed.

Kind Regards, Andreas

Hi Andreas,

Thank you so much for the reply and info; Much appreciated!  I turned up a CentOS 6 VM and the installation, using the RPM's provided on the 'downloads' page here, worked perfectly and was very straight-forward and easy.

The D/L page includes the Debian repositories (which is how I installed on Debian in the first place) but only installed v2.5.2.  I needed to get to the newest version (ELOG V2.9.2-2455) so I can setup authentication and SSH.  (And also so that when I am reading the manuals and documentation online, provided at this site, that is makes sense to the version I am running.)  I also want to add that it seems that I am able to bring over my older 2.5.2 logbooks to 2.9.2 without any problem.  To do that, I just rsync'd the logbooks from old to new.  They seem to work just fine under the newest version.

I'm good to go.  Anyone looking for an easy deployment of ELOG, especially in a production environment, I can attest that the RPM's provided here make it very simple to deploy on Cent6.

Thanks!