Maybe you should NOT use LO 3.5...

I was editing a simple document for about 30 minutes and then some of the figures started to disappear with an error message stating that the graphic could not be read. I saved the file (under a new name) and reopened. Those graphics still are not there. Opened the original and it is still there.

Is anyone else using LO 3.5? If yes, have any of you seen this? If not, beware!

I am using 64-bit Fedora 16 and I downloaded this from the web site. Guess I need to file a bug report. I also guess that I need to restart the editing process using something other than LO 3.5.

Hi :slight_smile:
Yes, we had this under 3.4.x too.  Apparently it has been a problem since about OOo 2.1 or something.  People have been trying to keep a back-up copy of images in a separate zip-file alongside the guides somewhere just in case it happens.  The zip-files have only been done for a few guides tho.
Regards from
Tom :slight_smile:

As Tom said, it's an intermittent fault that's been around for ages
and seems to affect files randomly (or at least in no reproducible
way). There is an existing bug report, though I don't recall the
number.

--Jean

I had not seen this behavior in a very very long time. I was under the impression that the issue had been fixed a very very long time ago. I have not had good luck with LO lately. Want to search outstanding bugs and such, but the web site appears to be not responding. It was a problem last night as well.

Hmm, well, I was able to find a bug that sounds about right...
https://www.libreoffice.org/bugzilla/show_bug.cgi?id=33393

Dropped a bunch of time filing bugs and such.

Time to eat some dinner then think about sleep. Will continue my pursuit later.

You've been lucky. Several of us encountered the problem with a few
files when using LO 3.4; I can't recall whether it occured with 3.3,
but I think so. My suspicion is that there is something weird in some
of our files ... some of which originated in OOo1.x, OOo2.x, or even
MSWord ... and that weird something triggers the observed behaviour,
sometimes, for some people, under who knows what conditions. Of
course, that's just speculation.

--Jean

Hi :slight_smile:
I suspect there are a dozen or so highly specific combinations of factors that cause the bug.  In which case it would be impossible to isolate the different sets from each other if there was any overlap.

Lets say that B=3 causes the bug but only when A is set to 3 or when C is set to 5 but if D=8 when A=3 then the problem doesn't happen unless ....  Add in that we don't know whether A is the OS in use or the one previously used with the file, or a version of java or phase of the moon or something to do with chicken bones.  Hmm, we can probably discount those last 2.  If devs haven't been able to solve it in several years then maybe the problem needs to be approached differently or maybe it will become obvious during code-clean-up or just magically vanish.

Regards from
Tom :slight_smile:

Andrew,

if you could reproduce such behavior, it'd be of great value for the
developers I'd suppose. Not constantly reproducible failures are very
difficult to pin down. I'd suggest to write a bug (or contact the qa list
first) and then to try to reproduce the failure with debugging enabled. I'm
not an expert but IMHO this could help to find the root cause of this
occasional figure loss. It also would help to save the original document
(before the loss) for investigations on its (maybe unvisibly broken)
structure, and of course, the broken document, too.

Just an 2¢ opinion

Nino

Blew a bunch of time on it already. I intend to spend a few hours trying to reproduce with my original files... Last time I had a problem like this it was tracked down to a strange combination in the way that the image was embedded as a character inside a frame (and that has long since been fixed).

My original speculation was that this was introduced in the code cleanup (which is totally false if this existed in OOo just before the split). I track some of the developer patches when I have time and many of the changes are very far reaching with some very tricky changes.

If I can reproduce I may spend some time trying to find where things go awry, but for this one I would need to poke some developers. last issue I found I was able to track it in the code and I ended up emailing with the developer that inadvertently introduced the bug. Although we disagreed with the final solution, a solution is in the current LO version (I checked that it made it into this release). This particular bug would exist in code where I have never looked.

Building LO was much easier than last time I tried to build OOo, which was many years back.

I am now able to cause it to happen every time in the latest version of LO (so I found one magic combination). This magic combination does not cause a problem in OOo. Certainly there may be more than one cause and/or trigger. I have some hope that if we find one cause and trace what it does, it may provide incite into other combinations (or even prevent the other combinations from causing a problem).

Spent a few hours today making sure that I can reproduce the problem. I also spent time poking around the code.

I have seen little to no evidence that anyone is actively pursuing this particular bug (which means almost nothing), but I do see that it is assigned to someone.

For the trigger that I found, it occurs when auto-save writes stuff out, but it does NOT occur if I write the file myself. I have a really long shot crazy notion or two, but I know so little about the internals that I expect that I am dead wrong.

Hi :slight_smile:
Good work!  Getting 1 cause out of the way would be great.  Sounds likely to be the commonest cause of the problem for this team.  Have you already posted a bug-report?
Regards from
Tom :slight_smile:

I jumped on to an existing report

https://www.libreoffice.org/bugzilla/show_bug.cgi?id=33393

Looks like exactly the same thing. Found a couple of reports marked as duplicate