6.12 Unstable

[jquest]

Hi,

I'm running 6.12 on Ubuntu 9.04, AMD64, up to date.

I've noticed a number of things with version 6.12:

1. Calibre seems to exit often when doing large numbers of convert to mobi operations, typically from lit to mobi.
2. after one of these crashes it's difficult to get calibre to start again, without manually going into /tmp and removing all calibre* and py* files and directories.
3. Calibre seems to hang often during a convert process, again converting from either pdf or lit files - typically it seems one or more files just 'won't convert'. If I exit calbre, clear the temp dir as in #2 above, and then retry the hanging conversion, often it will go through quickly.
4. It seems as though the 'Show Job Details' button no longer gives you any 'live' status - details will only show up when a conversion job has been completed.
5. Importing groups of files into calibre seems to often make the program exit - most of the groups are combinations of pdfs and lit files.
6. When I try to view a pdf file in calibre, nothing happens. Viewing the file using the external 'evice' browser works just fine, but calibre doesn't seem to be able to view them.

Is there a specific way that I should be running the system to give you debug outputs to help you with the issues? I figure if I am having them, others might be too, and Calibre is too great a program to have these instabilities.

Cheers,

Ned

[kovidgoyal]

You're using the calibre binary installer? Or did you build from source?

[jquest]

The binary installer

[kovidgoyal]

I've fixed a couple of memory leaks that should take care of 1.


For 2 and 3, try running calibre as

calibre-debug -g

from a terminal

4. the job output is stored to disk now, so buffering of disk IO will cause this.

5. Again calibre-debug -g

6. This will a dependency problem when running the binary installer (it has its own bundled libraries which probably conflict with the system ones preventing evince from running.

[jquest]

great, I'll start running it that way and send debug traces as they occur.

Cheers,

Ned

[jquest]

OK, first time I got to try this new method.

I'm running it from a shell script as follows:

rm -rf /tmp/calib*
rm -rf /tmp/py*
bash -c 'calibre-debug -g > calibre.log 2> calibre-error.log' >calibre-bash.log 2>&1

it created two logs this time - calibre.log, & calibre-error.log

They are attached to this post.

I was importing a bunch of pdb files when this happened.


Thanks.

[jquest]

Now I have removed the 'rm' lines from the script and started again.

The main screen starts to come up, but then it dies immediately.

The log is attached.

I'll put the 'rm' lines back now.

[jquest]

still won't come up.

Same log as list time, but I looked in /tmp and saw the attached two log files.

Hmm, now afterwards if I just run straight 'calibre' from the command line, it comes up fine.

maybe this problem is an artifact of running in debug mode after a crash?

yep, must be, since shutting down calibre after the non debug successful startup, then running the debug version seems to work ok.

[jquest]

I imported about 10 pdb files, all going well, then edited the metadata to fix some input, and then started a bulk convert of all of these 10 pdb files into mobi.

It ran for a bit, I think one got converted completely and then it died.

Attached is a tar file of the logs from both my shell script, and from the tmp directory.

I've also attached the logs from the 2nd run of bulk convert on the 9 remaining pdb files to mobi.

and one final run in the 3rd tar, I had converted maybe 8 files, choosing to convert them individually rather than in bulk, but as the number of pending jobs got over 10, then it choked again.

[kovidgoyal]

OK looks like a segfault when the worker process sends its result to the main process running the GUI. Next thing to do is run the gui with strace as follows

strace -f calibre

Also, I'd suggest waiting a couple of days until I release 0.6.13 as it has a few stability fixes that might take care of this.