Hi We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap). Any ideas - is this fixable? Thanks, Hilary.
+++ Hilary Greaves [2009-02-13 23:17 +0000]:
Hi
We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap).
Any ideas - is this fixable?
'Bus error' actually means something like you have run out of memory, or the software has segfaulted - i.e there has been a serious disaster at a low level. There may well be a corresponding software bug, but trying another computer would be informative if you have one to hand. Wookey -- Principal hats: Balloonz - Toby Churchill - Aleph One - Debian http://wookware.org/
Wookey wrote:
+++ Hilary Greaves [2009-02-13 23:17 +0000]:
Hi
We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap).
Any ideas - is this fixable?
'Bus error' actually means something like you have run out of memory, or the software has segfaulted - i.e there has been a serious disaster at a low level. There may well be a corresponding software bug, but trying another computer would be informative if you have one to hand.
Hils is struggling with the same dataset I am. She's running OS/X and gets a Bus Error. On Ubuntu (using the distro provided therion 0.5.1) I also get Segmentation faults with exactly the same test cases. I'd guess it's dereferencing a null pointer somewhere. Cheers, John
Wookey wrote:
+++ Hilary Greaves [2009-02-13 23:17 +0000]:
Hi
We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap).
Any ideas - is this fixable?
'Bus error' actually means something like you have run out of memory, or the software has segfaulted - i.e there has been a serious disaster at a low level. There may well be a corresponding software bug, but trying another computer would be informative if you have one to hand.
Hils is struggling with the same dataset I am. She's running OS/X and gets a Bus Error. On Ubuntu (using the distro provided therion 0.5.1) I also get Segmentation faults with exactly the same test cases. I'd guess it's dereferencing a null pointer somewhere. Cheers, John
On 14.2.2009, at 2:07, John Pybus wrote:
Wookey wrote:
+++ Hilary Greaves [2009-02-13 23:17 +0000]:
Hi
We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap).
Any ideas - is this fixable? 'Bus error' actually means something like you have run out of memory, or the software has segfaulted - i.e there has been a serious disaster at a low level. There may well be a corresponding software bug, but trying another computer would be informative if you have one to hand.
Hils is struggling with the same dataset I am. She's running OS/X and gets a Bus Error. On Ubuntu (using the distro provided therion 0.5.1) I also get Segmentation faults with exactly the same test cases. I'd guess it's dereferencing a null pointer somewhere.
Cheers, John
Just one strange think (I don't know if it is the source of error!). You use a bit strange definition of lakes there - areas water. They are defined by bottom wall and the water level is defined by another line wall -outline in. See attached picture 1. Standard way is something as you may see on picture 2. The line defined area is border and the wall clips the area inside the passage. In such case you may use "adjust horizontal" for line point to make the water level allways horizontal. Check the extension of Rabbit cave in samples. m.s.
Martin Sluka wrote:
Just one strange think (I don't know if it is the source of error!). You use a bit strange definition of lakes there - areas water. They are defined by bottom wall and the water level is defined by another line wall -outline in. See attached picture 1.
That was not a very good way of doing things, and we've fixed it (thanks for pointing it out), but I think we've found what was causing the bug, it was this: +line wall -close off + 320.0 -3625.0 +endline Somehow we'd managed to create a line with only one point. We'd actually noticed this before (and even isolated a test-case, but forgotten to send it to the list). That case is attached. There's little point to having a line with only one point -- it would very nice if therion provided an error message to warn about them. Cheers, John
On 16.2.2009, at 0:07, John Pybus wrote:
Somehow we'd managed to create a line with only one point. We'd actually noticed this before (and even isolated a test-case, but forgotten to send it to the list). That case is attached.
There's little point to having a line with only one point -- it would very nice if therion provided an error message to warn about them.
You may check your th2 files in plain text editor - check my mail from 2009.09.21- null lines xxxxxxxxxxxxxxxxxxxxxxxx check the th2 file for such null or single point lines: line wall endline line wall -subtype unsurveyed endline or line border -close on # originaly two point line, but closed, so there are two line points in identical coordinates 699.5 -2003.0 699.5 -2003.0 endline or :) line wall -close off 320.0 -3625.0 endline xxxxxxxxxxxxxx I made a simple grep patter to find such "null lines" in th2 files: I have try grep search/replace ^\s*(\w+) \w+\s+end\1 and it look it works Martin
There's little point to having a line with only one point -- it would very nice if therion provided an error message to warn about them.
I have found the problem in the source code (it is a null pointer usage). It will be fixed in the next version. Thanks for pointing out, S.
Hilary Greaves wrote:
Hi
We're trying to draw an extended elevation, and keep getting 'bus errors' on trying to compile the thconfig. One set of files that triggers an error is attached (this one just tries to draw a single scrap).
One thing to add about the test-case Hilary sent: If you remove the "-flip horizontal" from la_texa/elev_bottom.th2 then the survey compiles without a segfault (albeit giving a highly distorted passage as the scrap is the wrong way round). On a related note: We've been using -flip horizontal quite a lot as we combine two connected cave elevations which were not drawn the "same way round". To test various configurations we use "revise" commands in another file to add flips to selected scraps. What is not possible is using revise to remove a flip which was originally marked -flip horizontal. Would it be possible to add an option "-flip none" which could be used with revise to remove the flip? John
Martin was on to it. Correct the two water area 'drawings' and that fixes the problem. Looks like a nice cave when a loch model is made. The dataset as originally submitted caused Windows to close Therion and left Xtherion hanging. No error messages - just a crash. You guys clearly know more about revise, flip and extend that I, so I'll watch this space with baited breath. Bruce
Bruce Mutton wrote:
Martin was on to it. Correct the two water area 'drawings' and that fixes the problem. Looks like a nice cave when a loch model is made.
Yep, it's part of an 1100m (thus far) deep system.
The dataset as originally submitted caused Windows to close Therion and left Xtherion hanging. No error messages - just a crash.
Looks like the bad wall line on the water area is causing this problem, though ideally therion would report an error (and let us know which scrap is causing the issue) rather than crash entirely.
You guys clearly know more about revise, flip and extend that I, so I'll watch this space with baited breath.
I get the feeling that the extend elevation support in therion is not as solid as for other projections. We have had some other issues in the EE (whereas the plan of 7km of cave has gone smoothly). So far we've managed to work around them; expect more bug reports as we isolate the problems into test cases. John
participants (6)
-
Bruce Mutton -
Hilary Greaves -
John Pybus -
Martin Sluka -
Stacho Mudrak -
Wookey