Timeshift: Seeking fixes #155
Conversation
perexg
force-pushed
the
timeshift-master
branch
from
08d62e3
to
2205365
Compare
|
Did some tests and it seems working better than before ;-) @ksooo, shouldn't there be time-shift indicators present in the GUI? I thought that this was implemented in Jarvis but I can't see them. |
Cool. Could you please elaborate what "better" means. Especially, what does not yet work?
Core (incl. pvr.hts) support was implemented for Jarvis, but no Confluence UI. There was a lengthy discussion in the forum about that and people did not like what @FernetMenta proposed and "decided" to live with nothing instead of something semi-optimal (from their POW). I like what Fernet proposed and use it with my private Jarvis build for months, btw. => You may want to give https://github.com/ksooo/xbmc/commits/timeshift-confluence a try.
Sounds strange. No idea. Not my area of expertise, sorry. |
It's true, it's been like that since forever. |
|
Tested with tvheadend 4.0.8, skipping now works reliably. There's still some issues with fast-forwarding/rewinding but I haven't tried tvheadend 4.1 yet so that could be a non-issue too. I suggest we merge this. |
Also tested with 4.0.8. What shall I say? I was timeshifting and the first skip forward I did (no stop before), led to a hang of kodi. :-/ |
|
That's weird. On-demand disabled? |
|
Yep. Also RAM for buffer is disabled (if this makes a difference). But looking at the tons of changes and fixes @perexg currently makes on tvh master for the timeshift code and as he wrote that one should test against very latest tvh 4.1 I do not wonder that 4.0.8 does not really work. |
|
@perexg tvh master seems completely broken now, tvh will hang when using timeshift? |
|
Many timeshift related changes are in tvh master - things are stable in my testing environment. I might probably backport the timeshift changes to 4.0.x tree, but I'd like to test it more intensively before. |
|
@ksooo do you mind trying again? It never crashed/froze for me. If it does freeze, can you check with gdb where it does it? |
|
Not sure whether it is actually worth testing against 4.0.x before stuff got backported. |
|
Just a note: TVH v4.1-1299-g80a50ce seems to catch all major timeshift bugs. |
In combination with the pvr.hts fixes of this PR, I guess. Which Kodi version are you using for testing? Latest Jarvis beta? |
|
No, 15.2 and code from PR #144 . But it should be fine to use also this cherry-picked code from this PR. All changes just throws 'obsoleted' packets received between request / response states. And, the wait call in CHTSPDemuxer::Seek was really broken (missing variable initialization). |
|
Another "critical" timeshift bug fixed in v4.1-1301-g25b2cbc . |
|
Cross-linking TVH bug (Jarvis testers): https://tvheadend.org/issues/2744 |
|
@perexg can you rebase? |
|
Any news on this ? @perexg |
It's really a fatal error. If the predicate is not set to zero, then _PredicateCallbackDefault() function (see platform/mutex.h) gives true return value. It means that Wait() returns ASAP. This also means that m_seekTime must not be set to a zero value in the ParseSubscriptionSkip().
…ngeSpeed initialization
perexg
force-pushed
the
timeshift-master
branch
from
ea09b77
to
c832689
Compare
|
OK, rebased.. |
perexg
force-pushed
the
timeshift-master
branch
from
c832689
to
7449ad3
Compare
Cherry picked changes from PR #146 . I've not tested compilation on my dev platform (using 15.2), so little tweaking may be required.
Note: Test with TVH v4.1-1264-g0f069c7 and later.
Note2: Another bunch of timeshift fixes TVH v4.1-1276-ga27ed2e (mainly wrong refcounting and reworked on-demand mode).
Note3: TVH v4.1-1299-g80a50ce seems to catch all major timeshift bugs.