Re: [bug-gnuastro] Sorry, false alarm on make crashes

[Christopher Willmer]

Hello Mohammad,

Thanks so much for your feedback. Let me give a try to see if I can
find why these segmentation violations happen... As far as documentation -
you have done a stellar job in this. I will let you know how it goes.

Cheers,

Christopher



On Thu, December 1, 2016 08:40, Mohammad Akhlaghi wrote:
> Hi Christopher,
>
>
> On 12/01/2016 03:38 PM, Christopher Willmer wrote:
>
>> When I run configure with
>>
>>
>> ./configure --prefix=/home/cnaw/bin LDFLAGS=-L/home/cnaw/bin/lib
>> CPPFLAGS=-I/home/cnaw/bin/include
>>
>>
>> make runs perfectly; I got this tip from the manual. Is this a piece of
>>  information that would be appropriate to add in the INSTALL file ?
>
> I agree. It would be great if the INSTALL file could have this
> information. Currently, this file is not written by us (Gnuastro
> developers), it is imported from Autoconf (the program that generates the
> `./configure' script) during bootstrapping:
>
>
> https://www.gnu.org/software/gnuastro/manual/html_node/Bootstrapping.html
>
>
> So, instead of re-writing or modifying this file, I thought of fully
> describing this information in the manual instead:
>
> https://www.gnu.org/software/gnuastro/manual/html_node/Installation-direc
> tory.html (this is probably the page you checked, put here just as a
> reference.)
>
> We can work on a generic patch to this file (and discuss `LDFLAGS' and
> `CPPFLAGS') to send to the Autoconf developers. In this way, they can
> add it to the upstream `INSTALL' file and all programs that use Autoconf
> can benefit. It is indeed strange that these two important environment
> variables aren't discussed there.
>
> To be honest, after writing that thorough explanation in the link above,
> I don't have the time to do this now, if you can do it, or know anyone
> who can, it would be great.
>
>> Following on, make check  passes 44 tests, skips 1 (fitstopdf.sh,
>> because I am not using the latest gs) and fails 3, all with segmentation
>> faults:
>>
>>
>> FAIL: imgcrop/wcscat.sh
>> =======================
>>
>>
>> ./imgcrop/wcscat.sh: line 56: 12053 Segmentation fault      (core
>> dumped) $execname $cat $img --suffix=_wcscat.fits --zeroisnotblank
>> --numthreads=1
>> FAIL imgcrop/wcscat.sh (exit status: 139)
>>
>>
>> FAIL: imgcrop/radec.sh
>> ======================
>>
>>
>> ./imgcrop/radec.sh: line 52: 12150 Segmentation fault      (core
>> dumped) $execname $img --ra=0.99917157 --dec=1.0008283 --wwidth=0.3
>> --output=imgcrop_radec.fits
>> FAIL imgcrop/radec.sh (exit status: 139)
>>
>>
>> FAIL: imgcrop/wcspolygon.sh
>> ===========================
>>
>>
>> ./imgcrop/wcspolygon.sh: line 52: 12220 Segmentation fault      (core
>> dumped) $execname $img --wcsmode --zeroisnotblank
>> --output=wcspolygon.fits
>> --polygon=0.99980497,1.0001967:0.998378,1.0012267:0.9999766,1.0013217
>> FAIL imgcrop/wcspolygon.sh (exit status: 139)
>>
>
> These are the three tests that use WCS coordinates as inputs for
> ImageCrop (to crop regions in an image), the other tests use image
> coordinates. I guess the easiest way to pinpoint the source of the problem
> would be to use Valgrind. Could you try the following steps when ever you
> have some free time:
>
>
> 1. If you don't have Valgrind, could you install it? It must be
> available in the CentOS package manager.
>
>
> 2. Configure Gnuastro while adding the following options for no shared
> libraries, with debugging flags, and with no optimization so Valgrind can
> report the status accurately:
>
> $ ./configure --disable-shared CFLAGS="-g -O0"
> (this is also described in
> https://www.gnu.org/software/gnuastro/manual/html_node/Building-and-debugg
> ing.html )
>
>
>
> 3. These test scripts are located in `tests/imgcrop/XXX.sh' (from the
> top source directory), could you open them with a text editor and modify
> the last line so it starts like this (the `...' represent the rest of the
> line):
>
>
> valgrind --leak-check=yes $execname ...
>
>
> 4. Run `make' and `make check'.
>
>
>
> 5. The `tests/test-suite.log' file should contain the detailed Valgrind
> outputs.
>
>
> I am so sorry to ask for taking these steps (and taking your time). I
> have defined task #14258 to make it much more easier in the future.
>
> https://savannah.gnu.org/task/index.php?14258
>
>
> Thanks again for your patience with Gnuastro and reporting these issues,
> Cheers,
> Mohammad
>
>


-- 
Christopher Willmer
Steward Observatory, University of Arizona
933 North Cherry Avenue, Tucson AZ, 85721, USA
(520) 626-0805