Re: [Plplot-general] Test of plhist problem - it is empty again

[Alan W. I. <Ala...@gm...>]

On 2018-11-21 07:48-0000 Arjen Markus wrote:

> Hi Alan,
>
>> -----Original Message-----
>> From: Alan W. Irwin [mailto:Ala...@gm...]
>> Sent: Tuesday, November 20, 2018 10:21 AM
>>
>> Hi Arjen:
>>
>> That indeed looks like a good starting point for the failing test case.  But
>> fundamental debugging principles support the idea that another starting point that
>> uses plhist with success should be a big help to find out what is wrong with the
>> failing call to plhist.
>>
>> So I have the same question for you that I asked Sergey.  If you try using the code
>> from example 5 do you get the same rendered result as displayed at
>> <http://plplot.org/examples.php?demo=05>, i.e., plhist success?  Once you have a
>> succeeding test case and failing test case with plhist, then finding what the trouble
>> is with the failing test case should be completely straightforward.
>>
> I can confirm that the plot I get with x05c is very simular to that picture.
>
> However, I encountered a very strange and very unwelcome problem:
> When I added a few print statements to the plhist routine to find out what is going on, I had a coredump - I assume that is due to my print statements, though I just checked and did not see anything much that could cause that. To make sure no stray bits and pieces influenced this, I tried to rebuild the whole library and then, out of the blue!, CMake produced errors and failed to generate the makefiles. It complained for instance about files with a .cxx extension being unknown and no Fortran compiler being found.
>
> I have now started the comprehensive test script and that works fine! I am unhappily surprised about this. I will try to find out what on Earth is going on/wrong, but it may not be before tomorrow that I can really dig into this.

Your result suggested to me the possibility that plhist had a memory management issue
so I checked that possibility with valgrind, and the result was perfect, i.e.,

software@merlin> valgrind examples/c/x05c -dev svg -o test.svg
==5140== Memcheck, a memory error detector
==5140== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==5140== Using Valgrind-3.14.0 and LibVEX; rerun with -h for copyright info
==5140== Command: examples/c/x05c -dev svg -o test.svg
==5140== 
==5140== 
==5140== HEAP SUMMARY:
==5140==     in use at exit: 0 bytes in 0 blocks
==5140==   total heap usage: 604 allocs, 604 frees, 221,518 bytes allocated
==5140== 
==5140== All heap blocks were freed -- no leaks are possible
==5140== 
==5140== For counts of detected and suppressed errors, rerun with: -v
==5140== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

So from these results I think our plhist source code should be fine.
Of course, the perfect valgrind results here mean I have no clue at
all why you are running into intermittent trouble with plhist there.
Sorry I cannot help more, but I wish you the best of luck in figuring
out the issue on your platform.

Alan
__________________________
Alan W. Irwin

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________