[Libreoffice-bugs] [Bug 113937] Forms and reports hang

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=113937

--- Comment #37 from Ralph Peters <RPeters52 at gmail.com> ---
(In reply to Alex Thurgood from comment #35)
> (In reply to Ralph Peters from comment #34)
> > Well, no response so far - almost a month since my first post on this
> 
> That is a rather unfair assessment given that several people have responded
> to you over the weeks.
> 
> The fact of the matter is that you have a fairly large, complex embedded
> hsqldb database, and you  can not, for perfectly valid reasons, provide us
> with a copy of the ODB. This means we are pretty much limited in where we
> can go with this.
> 
> The fact that Base sigsegv's all over the place is nothing new and is down
> mainly to JVM instantiation and memory management (most, but not all, of the
> SIGSEGV tend to occur in the JVM).
> 
> Two suggestions :
> 
> - try increasing RAM allocated for the Java thread when you open the ODB
> file - possibly, that might help stabilize things a little - in order to do
> this, you will need to change the memory settings in the .properties or
> .scpt file within the ODB itself, which means unzipping and opening those
> files in a text editor, making your changes, then saving them and rezipping.
> Best to to do that on a copy of the ODB to see how well that goes. Off-hand,
> I don't remember exactly which file needs to be changed, but this should be
> searchable via Google, perhaps even in the hsqldb documentation itself ;
> 
> - second suggestion that might help us is if you could construct an ODB file
> with a minimum set of tables/forms/relationships not containing any
> sensitive data and that you could provide us, so that we could test
> ourselves.

I chose my words poorly Alex.  I have received a number of replies since I
first posted the problem!  I should have said that I have only received
requests for more info.

On a major piece of C++ code that I was developing, I ran both valgrind and gdb
at least every week during development to check for "unseen" problems.  Of
course, I ran them whenever I had a problem.  And that kept me out of trouble. 
So,the SIGSEVs caused me a lot of heartburn.  I had not received any response
about the SIGSEVs, until now. 

Yes, it is difficult to debug via "long distance".  I have done only a little
java software, so figuring out how to increase the size of RAM and the mutex
stuff would take me quite a while. BTW, I thought that java was not supposed to
have memory problems?  Wasn't that the point?  But I am merely an old C++
hacker (and lisp and Fortran)

The bottom line is that splitting the database SEEMS to have solved the
problem.  I am using hysqldb 2.4  This "experiment" may provide you with some
more ideas about what the problem is.  I will notify you on this thread if I
see a problem.  I have spent a number of hours adding data with 1 to 2 forms
open and NO problems.  I will be adding a lot of data to the split database and
will have multiple forms open at one time.  This should provide a good test!!

BTW, I will not go back to the embedded database as the split one posts info to
the database during the session and is reported to  be less likely to lose
information when there are problems.  Or do I mis-understand what I have read??

So, again my apologies for getting a bit rude.  The task, as I laid it out was
very difficult!!   I became irritated when I could not go "back to work".

Thanks for your help and attention!
Ralph

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20171219/72d21bba/attachment.html>