Memory Leak?

[adm]

Since fresh flashing for version155/6+ I've noticed some strange behaviour on my Miraclebox Twin. The problems still exist with another couch flash of build 166. No extra plug-ins, no timeshift, EPG with around 2 dozen channels. The box fan left permanently on. No backup restore of anything previously set-up.

After being on for around 4 to 6 hours the box starts becomes sluggish and slow to respond to the remote. For instance, selecting the (graphical) EPG and the spinning VIX will appear for 5 seconds followed by the EPG being written to the screen slowly in another 5 seconds.

Leave the box on for another 2 hours and the EPG will take twice as long to appear. Selecting any menu item will result in the spinning VIX for 10 seconds.

Leave the box on for another 2 hours and it will start rebooting every 15 minutes. This is not a full boot. The front panel will still report the channel being watched, the TV screen shows the normal startup VIX photo splash screen and after 10 seconds the box will revert to showing the TV picture again. No error logs are produced and no error message on the screen.

If left for a lot longer the spinning VIX appears forever and the box stops responding to the remote.

The timespan for starting to see the problem is reduced by perhaps an hour if recording during the period.

Switching the box off/on from the mains restores full working order for another 4/5 hours until the box starts becoming sluggish again. This suggests that its not a heat problem.

Although I have no proof, it is symptomatic of memory leak problems I've seen on other equipment.

[judge]

Enable debug logs, restart & upload one the next time this happens.

[adm]

Miraclebox Twin Premuim
2 satellite tuners on 28.2
EPG via ABM for the UK Freesat channels.

Clean install of Apollo 166 and rescan for channels
Default VIX skin and no additional plug-ins.

I've not had the chance to leave my box on for a very extended period until today

Today its been on for 12+ hours and has reached the state where it is permanently displaying the spinning VIX and is totally unresponsive to the remote. I can however FTP to the box and retrieve the log file it is creating.

Currently this log file is 28,002,791 lines long and growing by the minute – so here it is :)

Luckily blocks of info are endlessly repeated so approx 28 ,002,791million lines can be deleted.


It looks like the problem is “< 32354.592103> WARNING: EPG Cache is corrupt (eventData::~eventData), you should restart Enigma!”

Before crashing the box did a 10 second display of the VIX photo splash screen. The front panel only showed the name of channel being watched and NOT “booting” as it does when normally booting up.

[adm]

attachment


42140

[birdman]

IIRC, if the EPG is corrupt you need to shut down the GUI, delete the current EPG data (/media/usb/epg.dat?) then restart the GUI.
Leaving the corrupt one there won't help.

[abu baniaz]

You can also delete the epg cache in the EPG menu. No need to stop Enigma from running first if you do it this way.

Menu > Setup > System > EPG > Load/Save/Delete > Delete EPG

[adm]

IIRC, if the EPG is corrupt you need to shut down the GUI, delete the current EPG data (/media/usb/epg.dat?) then restart the GUI.
Leaving the corrupt one there won't help.


However once crashed, switching off at the mains and switching back on again restores full operation including a fully populated EPG, presumably read from the USB drive. The error message of "EPG Cache is corrupt " suggests its the copy in RAM that has become corrupt. If I use the box, say, 4 hours a day for weeks and switching to deep standby in between I only see a very slight slowing down at the end of the sessions. And this is with extensive channel hopping by way of the graphical EPG. I don't see crashes until the box has been left on permanently for much, much longer.

The question is what is repeatedly corrupting the EPG Cache/RAM? Running out of memory the longer the box is left on??

I will delete the existing EPG on the USB - I also have another drive I can try. Unfortunately it may be a while before I can report back as I have no plans to be in front of the box for any length of time in the near future.

[birdman]

The error message of "EPG Cache is corrupt " suggests its the copy in RAM that has become corrupt.And it is that which will get written out to epg.dat when you shut the system down and re-loaded as it starts up again. (If that didn't happen then there would be no point in having the file at all.)

[adm]

Quick update

Having followed the advice and deleted the EPG I've just had another random 10 second reboot

The last entry in the log file is

< 87476.813159> [EPGC] start caching events(1430504130)
< 87477.109810> sdt update done!
< 87604.349128> WARNING: EPG Cache is corrupt (eventData::~eventData), you should restart Enigma!
*** Error in `/usr/bin/enigma2': corrupted double-linked list: 0x020f0f10 ***


Still points to something corrupting the EPG.

[Sicilian]

Moved to correct section.

[adm]

Update
Hades 07, USB stick install, HDD formatted, new USB stick to hold EPG data and complete manual fresh set-up. The box/image still has the same problems with a report that the EPG cache is corrupt when it finally crashes.

However, disabling the automatic save and automatic refresh options in the EPG settings menu seems to stops this behaviour/crashing. Perhaps if a refresh and save occur at the same time the cache becomes corrupt?

Question: how does the box maintain an up to date EPG whilst the box is on? Does it require the automatic refresh option to be set to make sure that changes to the programming schedule, such as overrunning programs and/or last minute cancellations/substitutions, are captured every x hours?