New beta/experimental OSD rendering resolution for HD300 ?

[drewg]

Originally Posted by drewg

For me at least, all it takes is opening (by hitting enter on the remote) a malore recording menu, then selecting close on the menu. If I repeat that a few times, I'm sure to lock up the box.

BTW, I have the "advanced" malore menus (with the options to mark things watched, archived, etc) active. Not sure if that makes a difference.

Drew

[_Demo_]

If you are able to reproduce this issue could you create a miniclient debug log. The instructions to create it are at

http://forums.sagetv.com/forums/show...671#post375671

Thanks

[Opus4]

Originally Posted by drewg

For me at least, all it takes is opening (by hitting enter on the remote) a malore recording menu, then selecting close on the menu. If I repeat that a few times, I'm sure to lock up the box.

Originally Posted by drewg

BTW, I have the "advanced" malore menus (with the options to mark things watched, archived, etc) active. Not sure if that makes a difference.

BTW: if you are referring to the pop-up dialog that you see after selecting an airing which can be in Basic or Advanced mode, that is just that: the pop-up dialog with options for an airing. It is a standard dialog used on any menu where you can select TV airings.

The "malore menus" are a series of customizable menus where the TV airings are listed in a 2 pane display with groups on the left & airing lists for a highlighted group on the right. It is available as part of the hidden extras.

I think you are referring to the standard dialog.

I have tried reproducing the lock you described, but have been unable to do so.

- Andy

[voidpt]

@ Demo & Opus4

Tested some between firmware 20101022 & 20101007. I have no problem forcing the problem on 1022. It seems to happen every single time I open a popup-play-dialog-box and choose to close it. It freeze's in the middle of trying to close. Here is a bad mobile phone picture of my TV at this point:
TV_OSD_freeze.jpg

What I've done now is try to do the same operation twice. One with 1022 and one with 1007. Going into Current Recording, selecting a recording, then pressing close on the pop-up dialog box. Then I've telnet'ed and saved sagetv_0.txt from HD300. And I've also saved a copy of sagetv_0.txt from my SageTV Server. Have uploaded these to your FTP site. You can find them under ./From voidpt/.

PS: For some reason my sagetv_0.txt log from SageTV Server has an entry "MiniClient GFX_ASPECT=" with 15 0x00 hex chars. Make sure you open it with a text editor that can handle that to see whole log. UltraEdit went into hex-view per-default.

PS2: The time references are 2h different in logs.

PS3: Have tried to revert back to Standard from Serenity theme. Same problem.

[_Demo_]

Originally Posted by voidpt

@ Demo & Opus4

Then I've telnet'ed and saved sagetv_0.txt from HD300. And I've also saved a copy of sagetv_0.txt from my SageTV Server.

Could you send us the messages files. Those should contain the useful information for this problem.

Thanks

_Demo_

[drewg]

Originally Posted by _Demo_

Could you send us the messages files. Those should contain the useful information for this problem.

Thanks

_Demo_

What messages file? Syslog does not seem to be running...

tango3[~]# find / -name 'messages*'   
/app/dcchd/directfb/include/directfb/direct/messages.h
tango3[~]# pgrep syslog

Drew

[drewg]

Originally Posted by voidpt

It seems to happen every single time I open a popup-play-dialog-box and choose to close it. It freeze's in the middle of trying to close. Here is a bad mobile phone picture of my TV at this point:
Attachment 11424

This is the exact same thing that happens to me.. I notice you've also got the "advanced" popup enabled. I was going to switch to the basic one, and re-try if I get a chance.

Drew

[gplasky]

Originally Posted by drewg

What messages file? Syslog does not seem to be running...

tango3[~]# find / -name 'messages*'   
/app/dcchd/directfb/include/directfb/direct/messages.h
tango3[~]# pgrep syslog

Drew

Did you:

To enable debug logging on the STP-HD200/300, both for standalone and extender modes: while in standalone mode (not connected to a SageTV server), do the following:
1) Go to System Information and enter '5309' to enable the secret extras. When enabled, there will be a light blue dot in the upper left corner of the menu.
2) Go to Settings -> Advanced Setup -> General, and enable Debug Logging. Do not enable REM or any other logging option unless asked to do so or else the logs will be much larger.

In extender mode, the 'messages' miniclient log will be created ON the STP-HD200/300, while the 'sagetv_0.txt' debug log will be created on the SageTV server.

Gerry

[drewg]

Originally Posted by gplasky

Did you:



In extender mode, the 'messages' miniclient log will be created ON the STP-HD200/300, while the 'sagetv_0.txt' debug log will be created on the SageTV server.

Gerry

No, but I manually started klogd/syslogd and got a messages file that way. I'll try the instructions above & see if its the same format.

I was hoping for some smoking gun before the "unbalanced irq" kernel messages pushed everything out of dmesg. Unfortunately, there is nothing unusual prior to those message..

Drew

[_Demo_]

The messages file are only created when using debug mode.

_Demo_

[drewg]

Originally Posted by _Demo_

The messages file are only created when using debug mode.
_Demo_

OK, I've enabled debug logging (and rem logging..) and I've got a messages file that you expect. I just uploaded hd300-logs.tgz to the directory gallatin on your ftp server. In the tarball you'll find the client and server logs.

One thing I found was that using the "Advanced options" pop up seems to be requirement. I switched to "Basic options" and I could not reproduce the bug. I noticed that voidpt also was using the advanced popup.

Again, this really looks like a kernel bug in the llad module. No matter what happens, an application crashing should not cause an unbalanced irq enable/disable. Sorry to fixate on this, but I'm a professional *nix driver developer, and this just screams that something is horribly broken. The kernel starts spewing these messages exactly when the lockup occurs. I don't know if it is the root cause, or just a side effect of sagetv crashing. But either way, there is a bug in that module that needs to be fixed. I wish it was GPL so I could see the source, but according to a modinfo that I cross-built, it is LGPL..

Best regards,

Drew

[voidpt]

Originally Posted by _Demo_

Could you send us the messages files. Those should contain the useful information for this problem.

Done. Going a little fast these days. Uploaded a ZIP with 3 variants on ./From voidpt/. Name of files gives it away:
- One with firmware 20101007 and advanced box going ok
- One with firmware 20101022 and advanced box going freeze
- One with firmware 20101022 and basic box going ok

Originally Posted by drewg

One thing I found was that using the "Advanced options" pop up seems to be requirement. I switched to "Basic options" and I could not reproduce the bug. I noticed that voidpt also was using the advanced popup.

Nice observation. You are correct. Tried "Basic" dialog box. And was also not able to reproduce the freeze. Back to "Advanced", instant freeze on next close.

[_Demo_]

Please try the new 20101026-0 beta firmware. It should also give more information if the problem still happens.

Thanks

[drewg]

Originally Posted by _Demo_

Please try the new 20101026-0 beta firmware. It should also give more information if the problem still happens.

Thanks

So far, so good. At least, I'm not able to instantly lock it up. However, the options popup closes very slowly. It pauses for a second or more partway through the closing animation. The pause seems to happen at roughly the same spot the lockup used to happen. Switching to "basic" options removes the pause.

I've left logs from the hd300 running for a while on your ftp server in gallatin/hd300.2.tgz. Just messages this time.

Thanks!
Drew

[voidpt]

Originally Posted by _Demo_

Please try the new 20101026-0 beta firmware. It should also give more information if the problem still happens.

Did a quick test before work this morning. Same result as drewg. It does not freeze (permanent). But does freeze for 0.x sec at same time as it hung permanently before (is noticeable). "Basic" no problem, like before.

Can create some 'messages' logs if of interest. Will also run the 1026 firmware for now. If there should be any other cases of freeze.

[_Demo_]

Originally Posted by drewg

I've left logs from the hd300 running for a while on your ftp server in gallatin/hd300.2.tgz. Just messages this time.

It seems the logs had already been rotated so it doesn't show any information about the hangs. If you could save one shortly after the hang we should be able to get a better idea of what happens.

Thanks

[drewg]

Originally Posted by _Demo_

It seems the logs had already been rotated so it doesn't show any information about the hangs. If you could save one shortly after the hang we should be able to get a better idea of what happens.

Thanks

Sorry.. I assumed that it would rotate the logs farther than .0, and that .0 was the oldest. I've just triggered the pause several times, and have collected logs.

Your FTP server won't let me in (complains of too many users), so I've attached a .zip file of the logs since the forum won't let me attach a .tgz. I create a zip file about once every 10 years, so if you can't read it, PM me, and I'll email you the tarball.

Anyway, it seems each pause when closing the popup correlates to a "Timeout on gfx commands" and a register dump in the log. There are several in my logs.

Drew

[_Demo_]

Please try to latest 20101027-0 beta firmware and let me know if it fix your timeouts.

Thanks

_Demo_

[voidpt]

Originally Posted by _Demo_

Please try to latest 20101027-0 beta firmware and let me know if it fix your timeouts.

Nice Haven't checked the logs, but the animation is smooth as it should be. Not a single trace of the freeze from before.

[TwistedMelon]

Originally Posted by drewg

Sorry.. I assumed that it would rotate the logs farther than .0, and that .0 was the oldest.

It does keep more logging info. Except, "0" is the NEWEST log file. Followed by 1, 2, etc. In other words, the ordered numbered files represent the log events going backwards in time.