OpenCPN Beta Version 4.8.10119

[stelian]

Originally Posted by bdbcat

I see the problem, I think. I corrected some probably related cm93 errors in github, anyway.
Please try the current github master.

Hi Dave,

Current git master silences the errors in the log file, but the bug is still there.

The problem is that, when you zoom in and try to load a CM93 cell, and this cell doesn't exist, all the cells of this zoom level are marked as unavailable.

See this (edited) example:

First I try to zoom somewhere, everything is ok, the file is present so the cell is displayed:
Attempting loadcell 4161074 scale E, sub_char 0 at lat: 48.6667/49 lon:358/358.333
filename: /path/to/03901020/E/04161074.E
Loading CM93 cell /path/to/03901020/E/04161074.E

Then I go somewhere else, this time the cell doesn't exist, so the keys shown (I've added debug traces for this) are added to m_noFindArray:
Attempting loadcell 4161076 scale E, sub_char 0 at lat: 48.6667/49 lon:358.667/359
filename: /path/to/03901020/E/04161076.E
Adding1 03901020/E/0
Adding2 03901020/E/0.xz
Adding3 03901020/e/0
Adding4 03901020/e/0.xz

Finally, I go again to another location, where the cell does exist !. But since the key is recorded into the array, the cell will be ignored (no 'Loading CM93...' message):
Attempting loadcell 4171073 scale E, sub_char 0 at lat: 49/49.3333 lon:357.667/358
filename: /path/to/03901020/E/04171073.E

$ ls -l /path/to/03901020/E/04171073.E
-r--r--r-- 1 stelian stelian 181061 janv. 21 21:43 /path/to/03901020/E/04171073.E

If you find it useful I can put a 350MB CM93 dataset for you to download (or I can even try to make a smaller set just to show the bug....)

I also see a completely unrelated issue regarding the size of icons on my hidpy screen (3200x1800), they are really tiny and hard to see. Oesenc charts icons are much bigger and confortable to use.

Stelian.

[Hakan]

Dave.. Stelian..
On Win10
I think I can confirm Stelian's issue.
Without further investigations so far I've a shot. The right part should have viewed the E/F parts as well.
The log says nothing but "loading CM93.........F and .C and .B
Håkan

[brucechun]

Hi Dave,

Most of the anchorage line problem were fixed. However, there is still anchorage line and restricted area lines missing. As attached, part of the anchorage boundary line was missing where two charts were joined.

4.8.1.119


4.4


Part of the restricted area line also missing.

4.8.10119


4.4


Thanks, best regards,

Bruce

[brucechun]

Dave,

It seems to me that the the size of AIS targets were out of proportion in compare to other features on a S-57 charts.

On a HP 608 G1 tablet (screen resolution 1536x2408)
The buoys were very small while the AIS targets were rather big already.


Object factor was changed to 5.
The buoys became reasonablely normal to view, but the AIS targets became huge.



The difference was not so big on a Surface 3 tablet (screen resolution 1920x1280)


Best regards,

Bruce

[rooiedirk]

Also having the CM93 Zooming issue.
When zooming in, close and restart O the detailed charts are loaded:
From the log:

12:47:26 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/F/04280015.F
12:47:26 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/E/04280015.E

When zooming the charts faile to load. From the log:

12:53:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/e/04280015.e
12:53:32 CET:    cm93chart  Error ingesting /home/dirk/Charts/CM93//03900000/e/04280015.e
12:53:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/e/04280015.e
12:53:32 CET:    cm93chart  Error ingesting /home/dirk/Charts/CM93//03900000/e/04280015.e

Pls not the different in normal and kapital settings. On a linux system like mine this does make a difference.
Also the double '//' might be a problem.

[rooiedirk]

After recompiling with the last CM93 commit I get in the log wenn zooming:

14:03:29 CET: Getting DefaultSystemLocale...
14:03:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/C/04140000.C
14:03:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/C/04140012.C
14:03:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/C/04260000.C
14:03:32 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/C/04260012.C

Note only C scale charts.
Then stopping and restarting O (at the last zoomed in zoomlevel:

14:05:57 CET: ChartSymbols loaded from /usr/local/share/opencpn/s57data/chartsymbols.xml
14:05:57 CET: Using s57data in /usr/local/share/opencpn/s57data
14:05:57 CET: Initializing Chart /home/dirk/Charts/CM93/
14:05:57 CET: CM93Composite Chart Root is /home/dirk/Charts/CM93//
14:05:57 CET: Loaded CM93 Dictionary from /home/dirk/Charts/CM93//
14:05:57 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/F/04280015.F
14:05:57 CET: Loading CM93 cell /home/dirk/Charts/CM93//03900000/E/04280015.E

It's all uppercase now, but still not correct.

Edit:
Maybe it does make a different, but I am using the last available CM93 set, wicht is twice the size of the previous one, and a double amount of files now 30403 files in 1710 folders and 1.8 Gb. Previous set was 19295 files in 910 folders and 1.3Gb
Edit2:
The same chartset does work fine using stable 4.8

[bdbcat]

Stelian/Dirk/Hakan...

I corrected the cm93 noFind array logic again. Some scrambled code cleaned up. The original intent was to cache the full path of missing cells, leaving off the common cm93 fileset prefix for performance reasons. That part was broken.

Background:
Some platform file systems do a terrible job of looking for a non-existent file. This entire idea is to avoid operating system/file system access where possible, by prior or learned knowledge.
The logic could probably be sped up even more if we change from a linear array search to a hashmap. We could use a clever hashing function which converts the reduced cell path name string to a non-colliding int type. We can get very fast hash lookup on int keys.

Thanks for digging into this.
Dave

[stelian]

Originally Posted by bdbcat

Stelian/Dirk/Hakan...

I corrected the cm93 noFind array logic again. Some scrambled code cleaned up. The original intent was to cache the full path of missing cells, leaving off the common cm93 fileset prefix for performance reasons. That part was broken.

Dave

Dave,

Works fine now, thanks!

Stelian.

[Hakan]

Dave.
CM93 fine again also here. Performance is comparable good as well. I'm of course not against even better but it has also been inferior to this.
Thanks
Håkan

[rooiedirk]

Dave,
Only a quick test, but looks OK.
Thanks
Dirk

[rooiedirk]

CM93 again,
On 'normal' latitudes it works well. Zooming on high latitiudes, aspecial where there is half background and half CM93 chart, is extremely slow. On my system I'm talking minutes with a CPU load of 100%
Dirk

[rgleason]

Originally Posted by rooiedirk

CM93 again,
On 'normal' latitudes it works well. Zooming on high latitiudes, aspecial where there is half background and half CM93 chart, is extremely slow. On my system I'm talking minutes with a CPU load of 100%
Dirk

Git fetch & compile. CM93 fast and efficient at 40 deg but up near Ellesmere and north it gets slower. In fact it hung up on me so opencpn was not responding see attached log. Then went back up and a was not quite so rambunctious, it is definitely slower and more fragile.

[rgleason]

I think it is having trouble getting oriented, once it latches on to something it renders pretty quickly, but there is blue circle spinning at the beginning.

It is definitely more pronounced far north or far south, with more black screens while processing. Noticed this over Antartic Penninsula and up north. I do get the blue circle when zooming in to mid latitudes but not black screens. Has only crashed twice though.

[bdbcat]

Bruce...

re:
"However, there is still anchorage line and restricted area lines missing."

I believe that I have resolved these problems with new code for next Beta.
Have you had any luck investigating the failed update on cell 505/503?
Still it looks like a true error in the update file. But I would like to be wrong on this, and find some other checker/viewer to confirm or deny the idea.


Thanks
Dave

[bdbcat]

Dirk/Hakan/Rick...

Improved cm93 at high latitude. Now we skip the NoFind logic if the array becomes too big, as it will in areas where cm93 coverage is very sparse.
Used locally, as for "normal" navigation, the array stays small, so we still use it.

Thanks
Dave