Beta Beta Test of Major New Features - Page 3

Starson17 · 05-16-2010, 02:45 PM

Quote:

Originally Posted by chaley

Probably as a result of no longer keeping incorrect paths. Before, it would recalc the metadata once, correcting the paths in the process. Now it won't do the recalc.

So the path separator issue was the cause of all the metadata issues in the cache?

Quote:

Will look for it.

I'm seeing it repeatedly (at least 4 times so far), but I can't force it to happen. I'm not sure what triggers it.

Starson17 · 05-16-2010, 03:23 PM

Quote:

Originally Posted by Starson17

I'm seeing it repeatedly (at least 4 times so far), but I can't force it to happen. I'm not sure what triggers it.

It's up to 6 or 7 times. BTW, this is completely new. I'm pretty sure I would have seen it before and I didn't, so it's probably in the newest code. Once partially disconnected, there's no option to fully disconnect, only to connect to a folder, which is what I do, then I can either continue (it seems OK, but I worry about having connected twice) or disconnect. Usually I disconnect and then reconnect.

I haven't tried this to the SD card on the USB reader. That's up next.

chaley · 05-16-2010, 03:25 PM

Quote:

Originally Posted by Starson17

So the path separator issue was the cause of all the metadata issues in the cache?

That, and also preventing world peace and being a prime contributor to global warming.

Quote:

I'm seeing it repeatedly (at least 4 times so far), but I can't force it to happen. I'm not sure what triggers it.

Looking at the code, it appears that if USB devices come and go, the device manager might make a query that the folder device isn't prepared to deal with.

Any chance that you are connecting/removing things?

Starson17 · 05-16-2010, 03:56 PM

Quote:

Originally Posted by chaley

Any chance that you are connecting/removing things?

No. However .... First, this is happening on the local drive connections, but so far, not on the USB/SD card.
Second, even though I'm not disconnecting any USB devices, I am running on a machine that has a mouse on a 15' USB connection. It could have an intermittent issue (it does if I connect the USB hub at the end of that 15' USB cable). However, the mouse doesn't drop out, and the problem isn't appearing (knock on wood) on the USB/SD connection.

I've still got your old source. I think I may be very slightly behind you now, and I still have the first set of source from you. Shall I continue on checking the USB/SD or would it help to see if this really is something new, by going back to the older code to see if I have the same issue with it?

chaley · 05-16-2010, 03:57 PM

I have committed a change (rev 5246) that forces the USB vendor ID of the folder device to what as far as I can determine are invalid values (0xffff). There are some pathways through the code that might (only 'might') cause partial disconnects if recognized devices appear while another device is connected. In addition, I have guarded the path normalization not to attempt normalization of None, not that I think this has happend.

The difficulty here is that for the menus to change, the DeviceManager *must* have signaled that the folder device has disconnected. I can find no reason for it to do that.

chaley · 05-16-2010, 03:59 PM

Quote:

Originally Posted by Starson17

I've still got your old source. I think I may be very slightly behind you now, and I still have the first set of source from you. Shall I continue on checking the USB/SD or would it help to see if this really is something new, by going back to the older code to see if I have the same issue with it?

By your description, this isn't a problem that occurs once per week, so I don't see any reason to chase rainbows with old code.

If it stops happening with the new code (by that I mean it doesn't happen once per hour, which is what appears to be the case now), then we can go forward.

Starson17 · 05-16-2010, 04:05 PM

Quote:

Originally Posted by chaley

If it stops happening with the new code (by that I mean it doesn't happen once per hour, which is what appears to be the case now), then we can go forward.

I'll try the new code. The HD disconnect wouldn't go 10 minutes, and usually not 5. Just to make sure we're on the same wavelength, by "HD" I mean browsing to a folder on the local internal hard drive - and the one I've been using is on the raid, near the Calibre main library. The SD/USB has been solid for nearly an hour, with no drops.

chaley · 05-16-2010, 04:17 PM

Another possibility: if *somehow* calibre's eject action is being triggered (how, I don't know), and if *somehow* the device refuses to eject and raises an exception, then it is possible for the device to remain partially connected.

I have eliminated this possibility in revision 5247. The folder device treats all ejects as disconnects, but does not in fact eject the device. This should remove the above possibility.

Starson17 · 05-16-2010, 04:25 PM

Quote:

Originally Posted by chaley

I have eliminated this possibility in revision 5247

I'm not sure if I'm on 5247 or 5246, but so far, so good - 10 minutes w/o a disconnect.

chaley · 05-16-2010, 04:29 PM

If you are using bzr from the command line, you can find out by cd'ing to the calibre directory containing src and running bzr log -l 3

The bzr GUI is very helpful in these situations.

Starson17 · 05-16-2010, 04:35 PM

Quote:

Originally Posted by chaley

The bzr GUI is very helpful in these situations.

I've been meaning to try the bzr GUI. I was on 5246. It worked great for 16 minutes and I decided to merge in 5247 - I'm testing that now.
It's been good for 11 minutes, too. I'm out of bugs. It's been fun. If I find any more you'll be the first to know. All bugs, annoyances, etc. discussed in this thread seem to have been smashed. Good job!

chaley · 05-16-2010, 05:03 PM

Still need to add the confirmation dialog

Starson17 · 05-16-2010, 05:07 PM

We're up to half an hour: it still looks good. There are no signs of any problems.

For completeness: I forgot that the absence of a delete warning for Device ebook deletes was discussed in this thread. That's unchanged. Other than that possible issue, I've got no more comments.

chaley · 05-16-2010, 05:17 PM

Quote:

Originally Posted by Starson17

We're up to half an hour: it still looks good. There are no signs of any problems.

Quote:

For completeness: I forgot that the absence of a delete warning for Device ebook deletes was discussed in this thread. That's unchanged. Other than that possible issue, I've got no more comments.

Committed in revision 5248.

Service is our middle name! We aim to please! No job too large or too small! Reasonable billing rates! No VAT for 1 week, so hurry with the orders! Send all payments to Mr Jones, Custodian for Dormant Accounts, Bank of Nigeria!

Starson17 · 05-16-2010, 05:26 PM

Quote:

Originally Posted by chaley

Committed in revision 5248.
Service is our middle name! We aim to please! No job too large or too small! Reasonable billing rates! No VAT for 1 week, so hurry with the orders! Send all payments to Mr Jones, Custodian for Dormant Accounts, Bank of Nigeria!

Wrap it and bag it and stuff it in a sack.
Send it off to Kovid before he comes back!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Copia beta test - Get Your Invite Now!	Nate the great	Announcements	34	11-18-2010 07:48 PM
Beta Test of new Kobo Driver	timlegge	Calibre	58	06-22-2010 07:37 AM
Beta test of major new features	chaley	Calibre	130	06-04-2010 08:52 PM
DR800 V2 Beta Firmware Battery Test	CoolDragon	iRex	29	05-11-2010 05:37 AM
Mobile Forum Beta Test Here!	Alexander Turcic	Announcements	12	03-10-2004 05:38 PM

05-16-2010, 03:57 PM	#35
chaley Grand Sorcerer Posts: 11,742 Karma: 6997045 Join Date: Jan 2010 Location: Notts, England Device: Kobo Libra 2	I have committed a change (rev 5246) that forces the USB vendor ID of the folder device to what as far as I can determine are invalid values (0xffff). There are some pathways through the code that might (only 'might') cause partial disconnects if recognized devices appear while another device is connected. In addition, I have guarded the path normalization not to attempt normalization of None, not that I think this has happend. The difficulty here is that for the menus to change, the DeviceManager must have signaled that the folder device has disconnected. I can find no reason for it to do that.

05-16-2010, 04:17 PM	#38
chaley Grand Sorcerer Posts: 11,742 Karma: 6997045 Join Date: Jan 2010 Location: Notts, England Device: Kobo Libra 2	Another possibility: if somehow calibre's eject action is being triggered (how, I don't know), and if somehow the device refuses to eject and raises an exception, then it is possible for the device to remain partially connected. I have eliminated this possibility in revision 5247. The folder device treats all ejects as disconnects, but does not in fact eject the device. This should remove the above possibility.

05-16-2010, 04:29 PM	#40
chaley Grand Sorcerer Posts: 11,742 Karma: 6997045 Join Date: Jan 2010 Location: Notts, England Device: Kobo Libra 2	If you are using bzr from the command line, you can find out by cd'ing to the calibre directory containing src and running bzr log -l 3 The bzr GUI is very helpful in these situations.

05-16-2010, 05:03 PM	#42
chaley Grand Sorcerer Posts: 11,742 Karma: 6997045 Join Date: Jan 2010 Location: Notts, England Device: Kobo Libra 2	Still need to add the confirmation dialog

05-16-2010, 05:07 PM	#43
Starson17 Wizard Posts: 4,004 Karma: 177841 Join Date: Dec 2009 Device: WinMo: IPAQ; Android: HTC HD2, Archos 7o; Java:Gravity T	We're up to half an hour: it still looks good. There are no signs of any problems. For completeness: I forgot that the absence of a delete warning for Device ebook deletes was discussed in this thread. That's unchanged. Other than that possible issue, I've got no more comments.

Advert

Advert