Expired programs are not being deleted with master/slave backends.

[Ian Dall <ian@…>]

When there are separate master and slave backends, recordings made on the slave are not deleted properly. They are moved to the "Deleted" recording group, but the files are not deleted and the database entries are not removed from the 'recorded' table.

Since this does not free up space, eventually all expirable programs end up in the "Deleted" group.

If the master backend is restarted the files are deleted and the database entries removed.

This appears to be introduced with commit f78f9992d754390fa42f109e5139b8eaf224d076

This appears to be because AutoExpire::SendDeleteMessages() records programs which have been moved to the "Deleted" group in deleted_set. Autoexpire::UpdateDontExpireSet() initialises dont_expire_set from deleted_set and AutoExpire::FillDBOrdered() skips programs in dont_expire_set when constructing expireList.

I can see what is wrong but it is harder to understand what is supposed to be happening and figure out the fix. This is not helped by some some variable and function names. dont_expire_set seems to be used to mean "don't delete" rather than "don't expire".

I THINK removing all references to deleted_set and initialising dont expire_set to the empty set would solve the problem.

This issue would be masked by what looks like another bug when there are no slave backends. If the recording hostname is the same as the master hostname, autoexpired programs are just deleted without first sticking them in the "Deleted' recording group.

[Ian Dall <ian@…>]

Patch to remove references to deleted_set and initialise dont_expire_set properly.

[beirdo]

Removing improperly set milestone.

[Jason <jason-devel-mythtv@…>]

This bug has been biting me as well. Is there any tracing, testing, etc. that I can do to help move this along?

[Ian Dall <ian@…>]

For information, I have been running with this patch since I submitted it (over six months ago) with no problems.

[davescammell@…]

biting me too, will try this patch and report.

[stuartm]

If including patches with your tickets can you please set the ticket Type correctly? We can usually close tickets with patches faster but they will usually go overlooked unless they have been properly categorised.

[davescammell@…]

Works for me. The patch only required about one line to be amended against MythTV Version : v0.26.0-36-g8c68560-dirty (26 fixes) and has only been done against the MBE but the SBE is now happily reporting:

mythbackend[29383]: I DeleteThread mainserver.cpp:95 (delete_file_immediately) About to delete file: /m/video1/mythtv/9700_20121128074200.mpg & similar for the associated png file.

and seems to be keeping the disk from filling up now. Sorry that took so long to report on, I've had build issues on the MBE.

+1 on the patch.Thanks.dave

[Ian Dall <ian@…>]

In e150a19030a81e3c099eff76b280b886122529b1/mythtv:

Error: Processor CommitTicketReference failed

GIT backend not available

[Ian Dall <ian@…>]

In 8a7892b1ca1ec307d32a78153b57e012a721e9d3/mythtv:

Error: Processor CommitTicketReference failed

GIT backend not available

[Ian Dall <ian@…>]

In e38e08fcfa8e23136f0effdd54cc91eb434efd67/mythtv:

Error: Processor CommitTicketReference failed

GIT backend not available