Octave 4.0.0 on Win7
Have a larger project I'm working on. I call 'profile on' 'profile off' in several sections to benchmark runtime.
Recently, I changed a large block of code and was running through bug checking. As the section had a large number of syntax errors (it was late), it would halt execution, I'd fix the code, and rerun until it hit the next one.
But, every 2nd time it woulld halt, Octave would crash. I thought it was a GUI bug, so I restarted with --no-gui, and it happened again. This time I was able to catch the message from the terminal window since the Windows crash message held focus. The syntax error is expected and obvious, but the application crash appears to be due to the runnnig profiler. Can't cut/paste but will try to transcribe (odd spacing deliberate):
-----------
Error 'LorR' undefined near line 253 column 10
error: called from
etafn>BoundCondIntegFunction at line 253 column 3
etafn at line 18 column 6
Yangmodel_output at line 20 column 10
YangmodelWhole at line 54 column 1
A s s e r t i o n f a i l e d !
P r o g r a m : c : \ O c t a v e \ o c t a v e - 4 . 0 . 0 \ b i n \ o c t a v e - c l i . e x e
F i l e : / s c r a t c h / j w e / s c r / m x e - o c t a v e / t m p - s t a b l e - o c t a v e / o c t a v e - 4 . 0 . 0 / l i b i n t e r p / c o r e f c n / p r o f i l e r . c c , L i n e 3 9 1
E x p r e s s i o n : c a l l _ t r e e & & a c t i v e _ f c n ! = c a l l _ t r e e
This application has requested the Runtime to terminate it in an unusual way. Please contact the applicaiton's support team for more information.
panic: Aborted -- stopping myself...
attempting to save variables to 'octave-workspace'...
save to 'octave-workspace' complete.
-----------
The workspace file seems pretty small. I can attach it if you think there might be somethign hiding in there.
The Yangmodel_output script calls 'profile on' at the start, and the syntax error occurs inside a function call to etafn (in it's own etafn.m file) Again, this crash only occurs on the second instance of an error halting execution. Not sure why that would be.
Making a minimal example for a bug rpt may be a bit difficult as I'm starting with quite a bit of a mess. Have been using profile for a while without this, and had been getting periodic application crashes but never noticed such a repeatable pattern. Happy to try a few things if it can narrow down the problem or generate a minimal xeample for a useful bug report.
I did a quick savannah search, nothing less than a year old seems directly applicable but it seems related to Rik's comment here:
But both of those seem to relate to calling profexplore. I definitely call 'profile on' within an m file (script not function if that matters). But I'm not calling profexplore when the crash occurs.
Nick J.