116 replies to this topic

[littlesat]

This is because this is the only place were ServiceEvent wasn't used............ in comparision with all other code...

[BuGless]

Why is it bad?

Because it is not "best practice" to have one optional plugin depend on another optional plugin.
There are things like libraries that are meant for sharing and using, and there are plugins. Everything a plugin needs, should preferably be present in the base system, not in some other plugin.

If remove self["ServiceEvent"] = ServiceEvent() in GME work fine again.
Everything worked perfectly for 2-3 days ago, before this(self["ServiceEvent"] = ServiceEvent() ) change.
I can not understand the reason for this behavior now.

Well, I added that to homogenise support for serviceevents across multiple plugins. I'll try and see if I can figure out why your way of accessing it from autotimer breaks things.

[Dimitrij]

Well, I added that to homogenise support for serviceevents across multiple plugins. I'll try and see if I can figure out why your way of accessing it from autotimer breaks things.

The problem is as follows:
1) plugin TMBD

baseEPGSelection__init__ = None
def EPGSelectionInit():
global baseEPGSelection__init__
if baseEPGSelection__init__ is None:
  baseEPGSelection__init__ = EPGSelection.__init__
EPGSelection.__init__ = EPGSelection__init__

def EPGSelection__init__(self, session, service, zapFunc=None, eventid=None, bouquetChangeCB=None, serviceChangeCB=None):
baseEPGSelection__init__(self, session, service, zapFunc, eventid, bouquetChangeCB, serviceChangeCB)
if self.type == EPG_TYPE_SINGLE and config.plugins.tmbd.add_tmbd_to_epg.value:
  def redPressed():
   from Plugins.Extensions.TMBD.plugin import TMBD
   from Plugins.Extensions.TMBD.plugin import KinoRu
   cur = self["list"].getCurrent()
   if cur[0] is not None:
	name2 = cur[0].getEventName()
	name3 = name2.split("(")[0].strip()
	eventname = name3.replace('"', '').replace('Х/Ф', '').replace('М/Ф', '').replace('Х/ф', '').replace('.', '')
	eventname = eventname.replace('(18+)', '').replace('18+', '').replace('(16+)', '').replace('16+', '').replace('(12+)', '').replace('12+', '').replace('(7+)', '').replace('7+', '').replace('(6+)', '').replace('6+', '').replace('(0+)', '').replace('0+', '')  
	if config.plugins.tmbd.profile.value == "0":
	 self.session.open(TMBD, eventname, False)
	else:
	 self.session.open(KinoRu, eventname, False)
  self["tmbdActions"] = ActionMap(["EPGSelectActions"],
	{
	 "info": redPressed,
	})

2)plugin EPGSearch

baseEPGSelection__init__ = None
def EPGSelectionInit():
global baseEPGSelection__init__
if baseEPGSelection__init__ is None:
  baseEPGSelection__init__ = EPGSelection.__init__
EPGSelection.__init__ = EPGSelection__init__
# Modified EPGSelection __init__
def EPGSelection__init__(self, session, service, zapFunc=None, eventid=None, bouquetChangeCB=None, serviceChangeCB=None):
baseEPGSelection__init__(self, session, service, zapFunc, eventid, bouquetChangeCB, serviceChangeCB)
if self.type != EPG_TYPE_MULTI and config.plugins.epgsearch.add_search_to_epg.value:
  def bluePressed():
   cur = self["list"].getCurrent()
   if cur[0] is not None:
	name = cur[0].getEventName()
   else:
	name = ''
   self.session.open(EPGSearch, name, False)
  self["epgsearch_actions"] = ActionMap(["EPGSelectActions"],
	{
	 "blue": bluePressed,
	})
  self["key_blue"].text = _("Search")

Works that that one of two things.
If enabled EPGSearch, that does not work TMBD.
Similar problems of GME.
I understand that this is my personal problems, but very sorry that all this has stopped working.
I spent a lot of time before it worked.
Four lines of code=ServiceEvent and all my work for the year can be thrown in the trash.

Edited by Dima73, 23 January 2013 - 20:43.

[littlesat]

In fact from scratch it was fault... So we could go two ways... One is fix the plugins... The other fix is go for revert....

Edited by littlesat, 24 January 2013 - 00:33.

[Dimitrij]

In fact from scratch it was fault... So we could go two ways... One is fix the plugins... The other fix is go for revert....

One is fix the plugins=unsubscribe from these functions
I spent four nights trying to find a compromise solution.
He is not.

Edited by Dima73, 24 January 2013 - 06:45.

[littlesat]

We see bulks of fragments of code.... but we do not see anything you changed or highlited were the real clue pops up.... So sorry I do really not understand if you find a fix or so...

[Dimitrij]

I can not fix in plugins.
Only remove those features.
I do not understand why ServiceEvent inclusion in the code GME and EPGSelection gives a strange effect.

Edited by Dima73, 24 January 2013 - 07:57.

[littlesat]

I do not understand either.... I'm also thought it should be harmless.... Probable it was the incorrect understatement. Did you try to remove it in MGE?

[Dimitrij]

If you remove those two patches:
http://openpli.git.s...7bd1c8b6316663d
http://openpli.git.s...3635d6c9614b0e8
Then everything works perfectly.
I think that you will run into problems because of these patches in the future.
Some plugins use a similar method, for example Partnerbox,EPGSearch...

[adri]

If you remove those two patches:
http://openpli.git.s...7bd1c8b6316663d
http://openpli.git.s...3635d6c9614b0e8
Then everything works perfectly.
I think that you will run into problems because of these patches in the future.
Some plugins use a similar method, for example Partnerbox,EPGSearch...

These changes also break the 'Single EPG' screen.
This used to have the service name on top of the window.
After these changes, the service name is no longer visible and you have to guess which channel's EPG you are seeing.

[BuGless]

If you remove those two patches:
http://openpli.git.s...7bd1c8b6316663d
http://openpli.git.s...3635d6c9614b0e8
Then everything works perfectly.
I think that you will run into problems because of these patches in the future.
Some plugins use a similar method, for example Partnerbox,EPGSearch...

Well, littlesat, revert them for now, I'll check the existing Partnerbox and EPGSearch plugins to see what they do and provide the patches again after they are conflictfree.
For the record though, I'd like to note that manually running the init code for a plugin from another plugin is not a good idea in general and that practice should not be copied. It's a case of "I jumped into the fire because they were doing it too".

[BuGless]

These changes also break the 'Single EPG' screen.
This used to have the service name on top of the window.
After these changes, the service name is no longer visible and you have to guess which channel's EPG you are seeing.

Ah, that might provide the clue we've been looking for. The "manual inheritance" probably uses a local ServiceEvent object and layers it on top of the GME object, which then conflicts with the new ServiceEvent in there.

[Dimitrij]

Let's just go back back changes and forget this topic.
Three evenings spent on the search for the causes.
That is a lot, I feel sorry time .

[BuGless]

Let's just go back back changes and forget this topic.
Three evenings spent on the search for the causes.
That is a lot, I feel sorry time .

I presume littlesat will back them out shortly. Forgetting the topic is not an option though, if we were to do that everything there is a problem, the structure of the code never improves.

[littlesat]

Done...

And in the name of Enigma2.... sorry for all the inconvenience

Probably we need to add Service and ServiceEvent doing the same and put comments next to it... As this might always break some plugins...

Edited by littlesat, 24 January 2013 - 18:47.

[Dimitrij]

Thank you very much...

[BuGless]

If you remove those two patches:
http://openpli.git.s...7bd1c8b6316663d
http://openpli.git.s...3635d6c9614b0e8
Then everything works perfectly.
I think that you will run into problems because of these patches in the future.
Some plugins use a similar method, for example Partnerbox,EPGSearch...

I tried reproducing the problem, but I couldn't. I.e. I'm running *with* the patches in, yet Partnerbox and EPGSearch seem to function just fine. TMDB I cannot load, the plugin from the repository says that there is a version conflict for some .pyo files. Can someone explain to me what buttons I have to press in which order to actually see the/a problem? I'm testing using the PLi-HD skin, BTW.

These changes also break the 'Single EPG' screen.
This used to have the service name on top of the window.
After these changes, the service name is no longer visible and you have to guess which channel's EPG you are seeing.

Is that using the PLi-HD skin? Because again, I do not see the problem there, it works the same with and without the patches. More specific instructions on how to reproduce this, or where to look, would be appreciated.

[Dimitrij]

TMDB:
1)libxslt_1.1.26-r7_mips32el.ipk
2)libxslt-bin_1.1.26-r7_mips32el.ipk
3)python-lxml_2.2.8-r1_mips32el.ipk
http://openpli.org/f...post__p__312215
+
http://openpli.org/f...attach_id=51019

Edited by Dima73, 25 January 2013 - 10:43.

[BuGless]

TMDB:
1)libxslt_1.1.26-r7_mips32el.ipk
2)libxslt-bin_1.1.26-r7_mips32el.ipk
3)python-lxml_2.2.8-r1_mips32el.ipk
http://openpli.org/f...post__p__312215
+
http://openpli.org/f...attach_id=51019

All four packages are not in the feed, correct? (Just checking, in case I'm overlooking something obvious).

[adri]

These changes also break the 'Single EPG' screen.
This used to have the service name on top of the window.
After these changes, the service name is no longer visible and you have to guess which channel's EPG you are seeing.

Is that using the PLi-HD skin? Because again, I do not see the problem there, it works the same with and without the patches. More specific instructions on how to reproduce this, or where to look, would be appreciated.

This is using the DMConcinnity-HD-M-mod skin.
This worked fine before the EPG patches and now doesn't show the channel name in the top left of the Single EPG screen.
The '<' and '>' buttons on the top right side are normally visible.
The argument 'it works in the PLI-HD skin' isn't valid, if the change breaks all other skins available from the 3-rd party PLI feed.
I don't have a screen short available, since I'm at work.
Since the patch has been reverted, I'll try updating over the weekend and see if the problem disappears.

Adri.