Re: [Talk-GB] Pigging potlach ...

2012-01-12 Thread Dave Stubbs
On 11 Jan 2012 22:08, Lester Caine les...@lsces.co.uk wrote:

 Lester Caine wrote:

 I know what areas of data I am working on.


 OK Just crashed it again ... but having done at least a couple of hours
work and saved MOST of it.

 So to log current state ...
 Flash area just went grey
 Bottom line has Transfering data from ecn.t0.tiles.virtualearth.net...
 Processors are quiet
 RAM at 35%
 Swap 0%

 Trying to close tab get
 This page is asking you to confirm that you want to leave - data you
have entered may not be saved and this leaves a track in the grey area.


Do you have multiple browser tabs open? There's a horrible flash bug on at
least 64 bit Linux that will happily freeze any and all flash instances
within your browser. The culprit is usually some stupid video advert in a
tab you're not even using. It seems to get more likely the more flash is
running.

Symptoms are all your flash areas going grey and unresponsive.

Only mitigation I found is use a separate browser for editing. If your
normal browser is Firefox, just fire up chrome for potlatch only to stop
the interference. I was actually using this in a non-potlatch context:
trying to watch f1 practice on iPlayer at the same time as surfing news
sites.

Dave
___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-12 Thread Lester Caine

Dave Stubbs wrote:


On 11 Jan 2012 22:08, Lester Caine les...@lsces.co.uk
mailto:les...@lsces.co.uk wrote:
 
  Lester Caine wrote:
 
  I know what areas of data I am working on.
 
 
  OK Just crashed it again ... but having done at least a couple of hours work
and saved MOST of it.
 
  So to log current state ...
  Flash area just went grey
  Bottom line has Transfering data from ecn.t0.tiles.virtualearth.net...
  Processors are quiet
  RAM at 35%
  Swap 0%
 
  Trying to close tab get
  This page is asking you to confirm that you want to leave - data you have
entered may not be saved and this leaves a track in the grey area.

Do you have multiple browser tabs open? There's a horrible flash bug on at least
64 bit Linux that will happily freeze any and all flash instances within your
browser. The culprit is usually some stupid video advert in a tab you're not
even using. It seems to get more likely the more flash is running.

Symptoms are all your flash areas going grey and unresponsive.

Only mitigation I found is use a separate browser for editing. If your normal
browser is Firefox, just fire up chrome for potlatch only to stop the
interference. I was actually using this in a non-potlatch context: trying to
watch f1 practice on iPlayer at the same time as surfing news sites.


OK I did have a lot less active last night ... and probably was looking up other 
data to cross check things at that point in the evening. I DID have flash 
disabled completely simply to get rid of the advertising crap, so I probably 
want seamonkey configured with that off again, and just use a firefox window for 
editing ;)


Thanks for the nudge ... exactly the sort of thing I was looking for!

--
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Andy Allan
On 11 January 2012 00:21, Lester Caine les...@lsces.co.uk wrote:

 Well I'm on SUSE11.3 64bit into an AMD quad core with 8Gb RAM and Seamonkey
 2.6.1
 Rock stable with everything else I run.
 I'll switch to Firefox on another machine when I have a little more time
 tomorrow night.

 Just pissed me off that I'd fixed the same block twice, but not managed to
 save any of the work :(

Hi Lester,

Sorry to hear that potlatch is freezing on you. Maybe you can give us
some more details? First off, the flash version number would be a
great help, it's the most likely thing to be a significant difference
between your machine and others. Seamonkey vs Firefox is less likely
to be a trigger, since both use the same plugin mechanisms to load
flash which runs potlatch.

Also, can you describe the freeze? Is it just the save button that
stops working, or is it everything? When it freezes is your machine at
100% CPU (on one core) or idle? (top may help identify the process
if it's burning CPU). Is there anything in particular that triggers
it? What state is it in when it freezes - are you just panning around,
have you just entered a tag, is there a feature selected etc? For
these, a screenshot of the frozen p2 might be useful for us.

As Richard (albeit fairly bluntly) said, we've not heard similar
reports from other people, but they might just be silently enduring
it. Any further help you can give us to get to the root cause would be
awesome.

Thanks,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread John Sturdy
On Wed, Jan 11, 2012 at 9:41 AM, Andy Allan gravityst...@gmail.com wrote:

 Also, can you describe the freeze? ...

 As Richard (albeit fairly bluntly) said, we've not heard similar
 reports from other people, but they might just be silently enduring
 it. Any further help you can give us to get to the root cause would be
 awesome.

The slowdown I notice is on a single-core 512Mb machine, and I haven't
looked into full system statistics for it but the behaviour is
consistent with thrashing virtual memory... after editing for a while,
drawing ways with the mouse gets very slow (almost freezes); I think I
found the browser process had got very large, but I can't remember.  I
just guessed that the code to search for whether the mouse was
currently on an existing node was looking through more stuff than
would fit in RAM, that was what the behaviour felt consistent with.
I can investigate further if you like, but are people concerned with
such small machines (other than as thin clients) these days anyway?
(It might be of concern for mappers in poor countries getting
second-hand machines, for example.)

__John

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Lester Caine

Andy Allan wrote:

On 11 January 2012 00:21, Lester Caineles...@lsces.co.uk  wrote:


Well I'm on SUSE11.3 64bit into an AMD quad core with 8Gb RAM and Seamonkey
2.6.1
Rock stable with everything else I run.
I'll switch to Firefox on another machine when I have a little more time
tomorrow night.

Just pissed me off that I'd fixed the same block twice, but not managed to
save any of the work :(


Hi Lester,

Sorry to hear that potlatch is freezing on you. Maybe you can give us
some more details? First off, the flash version number would be a
great help, it's the most likely thing to be a significant difference
between your machine and others. Seamonkey vs Firefox is less likely
to be a trigger, since both use the same plugin mechanisms to load
flash which runs potlatch.

Actually that was the reason for checking out the other machine ...
This one has Flash 10.3 r183 and no option to upgrade to 11.1 but the other 
machine has 11.1 r102
I need to work out if it's worth all the agro updating this machine from 
SUSE11.3 which is perfectly functional, while SUSE12.1 is annoyingly different 
in too many ways :(



Also, can you describe the freeze? Is it just the save button that
stops working, or is it everything? When it freezes is your machine at
100% CPU (on one core) or idle? (top may help identify the process
if it's burning CPU). Is there anything in particular that triggers
it? What state is it in when it freezes - are you just panning around,
have you just entered a tag, is there a feature selected etc? For
these, a screenshot of the frozen p2 might be useful for us.

The flash area just stops updating and if I switch away and back, then
it does not refresh the area. It knows something is active as it gives the 
pop-up if you try and switch away and moving that pop-up leaves a trail in the 
flash area.



As Richard (albeit fairly bluntly) said, we've not heard similar
reports from other people, but they might just be silently enduring
it. Any further help you can give us to get to the root cause would be
awesome.

It's the usual problem of no time ...
15 mins tidying up an area while I'm waiting for something else to finish is 
time usefully spent, but fire-fighting why something random is happening takes a 
lot longer :(

Hence asking the question ...

--
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Richard Fairhurst
John Sturdy wrote:
 The slowdown I notice is on a single-core 512Mb machine, and I 
 haven't looked into full system statistics for it but the behaviour 
 is consistent with thrashing virtual memory... after editing for a 
 while, drawing ways with the mouse gets very slow (almost freezes); 
 I think I found the browser process had got very large, but I can't 
 remember.  I just guessed that the code to search for whether the 
 mouse was currently on an existing node was looking through more 
 stuff than would fit in RAM, that was what the behaviour felt consistent
 with.

I think there's always been a memory leak there with P2; I've noticed that
sort of behaviour for months. I've only done minimal bits of delving into
it, but suspect it's a case of _either_ some sprites being left on the
display list (maybe the invisible sprites which P2 uses to define hit
zones for each way/node), or that the event listeners by which Flash
responds to mouse clicks (and other events) aren't being cleared properly.

Unfortunately it's not trivial to debug - I think this sort of thing is the
way that Adobe makes their money from Flash (i.e. you need to buy Flash
Builder for the profiling tools). But that's not to say it's impossible, and
I'd very much like to do so.

(Incidentally, one simple thing you can do to speed P2 up is to clear any
GPS traces you're no longer using - it's easy to forget that you have, in
effect, a whole bunch of 1000-node ways sitting around!)

 I can investigate further if you like, but are people concerned 
 with such small machines (other than as thin clients) these 
 days anyway?

Please do. I am interested, definitely - P2 should aim to run efficiently on
any machine :) (and it's not entirely altruistic, as one of the machines I
use regularly is a 2004 PowerPC Mac).

cheers
Richard



--
View this message in context: 
http://gis.638310.n2.nabble.com/Pigging-potlach-tp7174091p7175878.html
Sent from the Great Britain mailing list archive at Nabble.com.

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Andy Allan
On 11 January 2012 11:21, Lester Caine les...@lsces.co.uk wrote:

 It's the usual problem of no time ...
 15 mins tidying up an area while I'm waiting for something else to finish is
 time usefully spent, but fire-fighting why something random is happening
 takes a lot longer :(

Sure, I understand. Think of my list of things more as a if it ever
happens again, which hopefully it won't... rather than me asking you
to spend time actively trying to pin it down.

Cheers,
Andy

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread John Sturdy
On Wed, Jan 11, 2012 at 11:40 AM, Richard Fairhurst
rich...@systemed.net wrote:

 I think there's always been a memory leak there with P2; I've noticed that
 sort of behaviour for months. I've only done minimal bits of delving into
 it, but suspect it's a case of _either_ some sprites being left on the
 display list (maybe the invisible sprites which P2 uses to define hit
 zones for each way/node), or that the event listeners by which Flash
 responds to mouse clicks (and other events) aren't being cleared properly.

Something like one of those would be consistent with what I'm seeing.
It doesn't happen on a larger machine, so I don't think it's just a
matter of the hit zone algorithm scaling poorly.

 Unfortunately it's not trivial to debug - I think this sort of thing is the
 way that Adobe makes their money from Flash (i.e. you need to buy Flash
 Builder for the profiling tools). But that's not to say it's impossible, and
 I'd very much like to do so.

Another oddity I've noticed (also probably deep within Flash) is that
P2 sometimes either doesn't respond to a keypress but does to the
corresponding mouse click, or just responds much slower to the
keypress.  (I notice this with add in advanced mode, versus the +
key.)  But I guess that by its nature Flash is more mouse-oriented
than keyboard-oriented, so this may be inherent to Flash rather than
specific to P2.

 (Incidentally, one simple thing you can do to speed P2 up is to clear any
 GPS traces you're no longer using - it's easy to forget that you have, in
 effect, a whole bunch of 1000-node ways sitting around!)

I don't think that's what's happening here (I get the effect without
any GPS traces loaded).

 I can investigate further if you like, but are people concerned
 with such small machines (other than as thin clients) these
 days anyway?

 Please do. I am interested, definitely - P2 should aim to run efficiently on
 any machine :) (and it's not entirely altruistic, as one of the machines I
 use regularly is a 2004 PowerPC Mac).

Next time it does it, I'll run top and get the process size, memory
usage, etc, for a start.

__John

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Richard Fairhurst
John Sturdy wrote:
 Another oddity I've noticed (also probably deep within Flash) is that
 P2 sometimes either doesn't respond to a keypress but does to the
 corresponding mouse click, or just responds much slower to the
 keypress.  (I notice this with add in advanced mode, versus the +
 key.)

Yep, I get that too. In these cases Flash is sending the keypress to a UI
component that isn't listening for it. We'd ironed most of those out in a
previous P2, but unfortunately Flex 4.5 (which we recently moved to) seems
to have changed things here a little, and we're not catching them all
again.

cheers
Richard




___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-11 Thread Lester Caine

Lester Caine wrote:

I know what areas of data I am working on.


OK Just crashed it again ... but having done at least a couple of hours work and 
saved MOST of it.


So to log current state ...
Flash area just went grey
Bottom line has Transfering data from ecn.t0.tiles.virtualearth.net...
Processors are quiet
RAM at 35%
Swap 0%

Trying to close tab get
This page is asking you to confirm that you want to leave - data you have 
entered may not be saved and this leaves a track in the grey area.


'S' did not save the last bit of work ... but I probably expected that.

--
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


[Talk-GB] Pigging potlach ...

2012-01-10 Thread Lester Caine

OK how many of you are having trouble editing for more than 10 minutes?
I've lost as much work as I've done this evening with potlach just freezing :(
I had the same problem at the weekend, but put it down to finger trouble, know I 
know it is software.


--
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-10 Thread Richard Fairhurst
Lester Caine wrote:
 OK how many of you are having trouble editing for 
 more than 10 minutes? I've lost as much work as 
 I've done this evening with potlach just freezing :(
 I had the same problem at the weekend, but put it 
 down to finger trouble, know I know it is software.

If you _know_ it's software, then you must have isolated the fault and be
able to fix it! Awesome! Can't wait for the patch!

Ahem.

P2 hasn't changed in 13 days, except for one utterly tiny change one week
ago (to the logic in loading splitting GPS tracks). So if something changed
at the weekend, it's more likely to be your system than Potlatch 2. FWIW
I've not encountered any issues myself nor had reports of any.

cheers
Richard



--
View this message in context: 
http://gis.638310.n2.nabble.com/Pigging-potlach-tp7174091p7174357.html
Sent from the Great Britain mailing list archive at Nabble.com.

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb


Re: [Talk-GB] Pigging potlach ...

2012-01-10 Thread Lester Caine

Richard Fairhurst wrote:

OK how many of you are having trouble editing for
  more than 10 minutes? I've lost as much work as
  I've done this evening with potlach just freezing:(
  I had the same problem at the weekend, but put it
  down to finger trouble, know I know it is software.

If you_know_  it's software, then you must have isolated the fault and be
able to fix it! Awesome! Can't wait for the patch!

Ahem.

P2 hasn't changed in 13 days, except for one utterly tiny change one week
ago (to the logic in loading splitting GPS tracks). So if something changed
at the weekend, it's more likely to be your system than Potlatch 2. FWIW
I've not encountered any issues myself nor had reports of any.


Well I'm on SUSE11.3 64bit into an AMD quad core with 8Gb RAM and Seamonkey 
2.6.1
Rock stable with everything else I run.
I'll switch to Firefox on another machine when I have a little more time 
tomorrow night.


Just pissed me off that I'd fixed the same block twice, but not managed to save 
any of the work :(


--
Lester Caine - G8HFL
-
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php

___
Talk-GB mailing list
Talk-GB@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-gb