OpenCPN Version 2.1.0 Beta

[Ptizef]

Hi Dave and all

First tests :
1) performance problem
First of all , you are right Dave , the VC+ debugger slow down the program and it's better in normal condition , but the problem is still here .
- With cm93 it's normal and the probleme comes clearly from bsb . In quilting condition , the moving , zooming is slow , and not smooth and worst in the region where there are many bsb. In no quilting condition when changing position , it's take several seconds for the chart arriving in the center of the screen to be dispalyed (see shots "testa" taken after moving . In seconds , a chart will appear)

2) General
This version seams to work fine , except the bsb slow moving/zooming . The quilt control seams clear .
All that is very good !
I detected a problem : in the shot "test1" attached ,I placed the boat in the screen center and without moving , I zoomed until "test2" where I got an over zoom . the larger scale charts have not been opened . I exit and reopen without moving and now it's good , the right chart is opened (shot "test3")
In this case opencpn do not choice the right chart when moving

3) Crash :
I got no crash when moving .. zooming ... but I got several crashes (not systematically) when re-building charts database (see crash1,crash2) . This crashes happened clicking and moving immediately after build ,( may be too early ?)


Best regards
Jean Pierre
Sorry but I'll be far from a good wifi for a couple of days

[nohal]

Dave, Will,
I like the quilting bar more and more...

Originally Posted by bdbcat

Ptizef: Jean Pierre, I wonder about your performance issues. Is it in quilting only? BSB or cm93, or both? Have you tried Release build from VS tools? I find the Debug build can be slow, since it is un-optimized.
Anyone else seeing serious performance issues with quilting?

snail feels like a Ferrari in comparison with S57 loading in my debug build. It's roughly 10-20 times slower than in release. Otherwise the app is slower but usually not that slower.
What's a bit annoying is that it sometimes behaves differently in debug and release build. And as such we get to the bugs

1) The canvas out of the chart is not cleared on pan or zoom-out (when there's no underlying chart) - visible on canvas_not_cleaned.
2) In release build I get a crash while using mouse wheel for zooming (especially out) at which scale it happens depends on the speed of wheel turning. Not reproducible while using keyboard or clicking the icon.
3) While trying to reproduce 2 in debug, instead of a crash, the app hangs (zoomout_hang)
4) When I move (keyboard or mouse) to a place where probably no charts would be displayed, i get a crash (canvas_not_cleaned -> pan_out_of_screen_crash)

Pavel

[Ptizef]

Just a precision for my vc++ use :
when I say "debugging" , in fact I build in release ( don't work in debug I don't know why) but I run in debug means : menu debug/start debugging
Sorry I'm not very skilled
Jean Pierre

[manimaul]

I'm wondering if part of the perfomance issues is because of being zoomed out too far where whole charts are scaled down to the size of a stamp.

Perhaps we should automatically change chart scale within a set percentage of truescale so this doesn't happen. I'm thinking -.25x to +4x. Any ramifications for this?

This would clean up the display in this situation also.

[nohal]

More info:
#377 only reproducible if the IENC set is loaded alone, with CM93 world chart under it the problems are gone.
I can't reproduce Ptizef's performance problems while testing in release on Marco's NZ chart set. The chart switching takes just a fraction of a second. In debug I get the same hang as in #377 so maybe something on my side?

Edit: and I almost forgot another crash on DB rebuild. A couple of lines from Ptizef's one but probably the same thing...

Pavel

[manimaul]

Center pixel +limited scale range inclusion:

Right now the available chart bar is built simply by charts available on center pixel. What would happen if we also included charts within .25x to 1.5x truescale that exist in the entire window bounds?

Would this eliminate the problem of small scale charts close to center pixel not appearing?

Will

[champ8242]

I have to kill VS to continue

[champ8242]

Hi,

I have two points;
1) exception, picture 1, occurred after (at) closing OpenCpn (red X)

2) Picture 2 shows 1 ENC chart included on screen and active. If I hover over the green ENC bar I can see in the left upper corner a preview of the chart. Picture 3 shows after one zoom in 5 ENC charts on screen active. If I hover over the first green bar I get the preview, if I hover over the second and other green bars I don't get a preview in at all. Any purpose of of the preview picture at the first bar?

[champ8242]

Hi,

I deleted the opencpn.ini (on purpose) to start from scratch.

Bart

[champ8242]

after removing all the charts from quilting and zoom-out a couple of times.

Bart
PS, I like the red X in the quilting bars

[cagney]

Dave,
I'm still chasing my "exclusive" bug. Not as fun as serving you with moles to whack.
I have had similar crashes yesterday, but no "nPoints" ....

While chasing "nPoints" strings in the log I notice a few things. I have a lot of lines like this one when loading ENCs

06:25:50 PM: Error: Failed to convert file "/opt/Navigation/ENC/US/US4AK5KM/US000526.TXT" to Unicode.

and the odd one like this

06:29:20 PM: CPL ERROR 0: S52plib:_renderCS(): ERROR no conditional symbology for: OBSTRN05)

Checking on tmercs, a merc chart showed incorrectly as active when viewing tmerc, but of of course not showing in the quilt. Dump 1.

Loading both raster and S57s for the waters between Puerto Rico and the USV, was interesting, a raster showed up as a thumbnail for a vector chart, and vice versa.
Dump 2 & 3.

Dump 4 is the now well known crash when starting a reload of the chart database in quilt mode and then hitting F9.
All testing done with:Version 2.1.0 Build 504a.

Thomas

[nohal]

Hi Thomas,

Originally Posted by cagney

06:29:20 PM: CPL ERROR 0: S52plib:_renderCS(): ERROR no conditional symbology for: OBSTRN05)

This one's mine - resulting from the new symbolization instructions compiled from various versions of standards. If you find more stuff like this, please post them (if you can isolate on which chart/position they happen, even better but not necessary)

Pavel

[cagney]

Pavel

I suspected that this one was up your lane...

Below are a number of lines that appears just before the error line in different places:

07:01:36 PM: Initializing Chart /opt/Navigation/ENC/US/US5PR52M/US5PR52M.000

07:09:01 PM: Initializing Chart /opt/Navigation/ENC/Hungary/ENC_ROOT/1H5D1520.000

10:28:06 PM: Initializing Chart /opt/Navigation/ENC/Caribbean_Sea/INT_4025_NAppPC/CO200008.000

So the error occurs when initializing any of these three charts. The last one is a Columbian chart. Download link towards the end of this page:Installing Charts | Official OpenCPN Homepage

Thomas

[cagney]

Dave,
Here is what you are looking for.
Good Luck!

Thomas

[cagney]

Dave
I just had an unusual crash.
I had a lot of Caribbean charts loaded, cleared that and loaded(not using quilting) CM93 and the Hungarian Inland ENCs. Zooming in on the river charts and then changing to quilting ....... made OpenCPN exit cleanly!!!! No crash.
I think this was what activated your special log-entries.
When trying to restart, I get an immediate crash(see dump). I deleted the opencpn.conf (saved a copy) and then restarted normally(the vector chart was actually gray instead of black). All details below.
Thomas